The emergence of drug-tolerant persister (DTP) cells poses a major obstacle to durable responses to EGFR tyrosine kinase inhibitors (TKIs) in EGFR-mutant non-small cell lung cancer (NSCLC). However, how metabolic programming orchestrates this transient and reversible adaptive resistance remains largely unexplored. Here we report that proline dehydrogenase 1 (PRODH) promotes the persistence of osimertinib tolerance in NSCLC by increasing mitochondrial respiration in an enzyme activity-independent manner. Mechanistically, PRODH facilitates the recruitment of sirtuin-3 (SIRT3) to mitochondrial transcription factor B2 (TFB2M), resulting in the deacetylation of TFB2M at lysine 127 and lysine 153. This post-translational modification is indispensable for the mtDNA-binding capacity of TFB2M and subsequent initiation of mitochondrial transcription. Loss of PRODH, or increased TFB2M acetylation, impairs DTP formation and delays tumor relapse. Collectively, these data reveal a previously unrecognized role for PRODH in linking mitochondrial metabolism to drug-tolerant persistence, suggesting that targeting PRODH or TFB2M deacetylation may offer a promising strategy to overcome osimertinib resistance and suppress tumor relapse in EGFR-mutant NSCLC.

Mutations in isocitrate dehydrogenase (IDH) are frequently observed in several aggressive malignancies, including gliomas, chondrosarcomas, and acute myeloid leukemia. Mutant IDH enzymes exhibit neomorphic activity that produces the oncometabolite D-2-hydroxyglutarate (D-2-HG), leading to widespread DNA and histone hypermethylation. This aberrant epigenetic state remodels chromatin accessibility, thereby compromising both DNA repair and replication. However, the mechanisms underlying IDH mutation-induced DNA replication stress and the therapeutic potential of this vulnerability remain elusive. Through a whole-genome CRISPR screen, we defined differential genetic dependencies between IDH1 wild-type and mutant cells. We identified the G2/M checkpoint kinase PKMYT1 as a selective survival factor for IDH1-mutated cells. Pharmacologic or genetic inhibition of PKMYT1 profoundly suppresses the growth of IDH1-mutated tumor cells in vitro and in vivo. Mechanistically, IDH1-mutated cancer cells exhibit profound replication stress, characterized by reduced nucleotide incorporation, slowed replication fork progression, and delayed S-phase entry. D-2-HG elevates H3K9 trimethylation by inhibiting KDM4A, thereby impeding DNA replication initiation and fork progression. PKMYT1 blockade alters cell cycle timing, forcing unscheduled mitotic entry with incompletely replicated genomes, resulting in elevated DNA damage and mitotic catastrophe. Collectively, our findings demonstrate a mechanistic link between IDH1-driven epigenetic dysregulation, replication stress, and heightened reliance on the G2/M checkpoint. Our findings propose PKMYT1 inhibition as a therapeutic strategy that exploits the intrinsic replication deficiency of IDH1-mutant glioma.

The prognosis of patients diagnosed with relapsed/refractory acute myeloid leukemia (R/R AML) remains poor, and allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a curative strategy. An integrated model incorporating both genetic and clinical features to inform clinical decision-making in the HSCT setting for R/R AML patients is still lacking. In the current study, 301 R/R AML patients who underwent allo-HSCT at Peking University People's Hospital from 2018 to 2023 were included in the derivation group. A total of 297 R/R AML patients from 11 other HSCT centers in China were included in the external validation group. A total of 54 R/R AML patients who underwent allo-HSCT at Peking University People's Hospital in 2025 were assembled as a prospective validation cohort. A comprehensive prognostic model combining clinical factors and genetic abnormalities, including disease status before HSCT, patient and donor age, FLT3-ITD, TP53 mutation, KMT2A rearrangement (KMT2Ar) and complex karyotypes, for R/R AML patients treated with allo-HSCT was developed. Patients were stratified into four risk groups: low-risk, intermediate-low-risk, intermediate-high-risk, and high-risk. In the internal validation cohort, the 1-, 2- and 3-year time-dependent area under the curve (AUC) values of overall survival (OS) were 0.768, 0.795 and 0.799, respectively. In the external validation cohort, the 1-, 2- and 3-year time-dependent AUC values of OS were 0.676, 0.729 and 0.786, respectively. The 1-year AUC of the prospective validation cohort reached 0.79. This model could refine risk stratification in R/R AML patients undergoing allo-HSCT, and potentially inform combination therapy strategies and post-transplantation monitoring.

Cancer cells adapt to microenvironmental cues through the coordinated integration of metabolic reprogramming, epigenetic remodeling, and immune regulation. An emerging central regulator of this adaptive network is coordinated nicotinamide metabolism, which controls both nicotinamide adenine dinucleotide (NAD) biosynthesis and cellular methylation potential. Within this network, the nicotinamide-metabolizing enzymes nicotinamide phosphoribosyltransferase (NAMPT) and nicotinamide N-methyltransferase (NNMT) govern the metabolic fate of nicotinamide (NAM), thereby coupling energy metabolism to epigenetic regulation and adaptive cellular responses. NAMPT, the rate-limiting enzyme of the NAD salvage pathway, sustains intracellular NAD pools required for redox homeostasis, mitochondrial function, DNA repair, and stress responses. Conversely, NNMT diverts NAM from NAD biosynthesis by catalyzing its methylation while consuming S-adenosylmethionine (SAM), thereby creating a methylation sink that remodels chromatin organization and promotes transcriptional plasticity. Beyond intracellular metabolism, coordinated nicotinamide metabolism extends to the tumor microenvironment through extracellular mediators, including extracellular NAMPT (eNAMPT), the emerging extracellular functions of NNMT (eNNMT) and its metabolite 1-methylnicotinamide (1-MNA), which regulate stromal remodeling, immune cell function, chronic inflammation, and immune evasion. Collectively, these complementary intracellular and extracellular activities establish a multilevel adaptive network that integrates metabolic homeostasis, epigenetic remodeling, and immunometabolic communication, thereby promoting tumor plasticity, therapeutic adaptation, and resistance. Here, we propose a conceptual framework in which NAMPT and NNMT function as complementary regulators of coordinated nicotinamide metabolism operating across interconnected intracellular and extracellular levels. By integrating metabolic, epigenetic, and immune crosstalk into a unified model of tumor adaptation, this perspective provides new insights into tumor progression and therapy resistance while highlighting context-dependent opportunities for mechanism-based combination therapies in cancer.

Glioma progression is driven by both tumor-intrinsic genetics and tumor microenvironment (TME) heterogeneity. Although tumor-intrinsic heterogeneity has been widely explored, the understanding of the isocitrate dehydrogenase (IDH)-associated TME in glioma remains less clear. In this study, therefore, we integrated bulk and single-cell transcriptomics, spatial transcriptomics, and proteomic data from glioma tissues and patient blood samples to characterize TME composition, subtype-associated molecular patterns, and their prognostic relevance. The data showed that IDH-wildtype (IDH-WT) glioma exhibited complicated TME, enriched in pericytes and immune cells, whereas IDH-mutant (IDH-MU) glioma displayed a TME dominated by oligodendrocytes and neurons. From a TME-related geneset, we identified four subtype-associated gene patterns: 179 genes in IDH-WT (WT-179) associated with extracellular matrix remodeling and angiogenesis; 68 genes in IDH-MU (MU-68) linked to neuronal and synaptic pathways; 30 genes in IDH-MU with chromosome 1p/19q-non-codeletion (M-N-30) enriched in RNA/protein binding; and 83 genes in IDH-MU with chromosome 1p/19q-codeletion (M-C-83) involved in neurodevelopmental processes. Single-cell transcriptomics profiled contributions of pericyte, myeloid, and lymphocyte subpopulations to these patterns, revealing distinct developmental trajectories and functional programs across glioma subtypes. Glioma tissue and serum proteomics detected WT-179-associated proteins and identified POSTN, LTF, PLA2G2A, and ANXA1 as candidate protein markers associated with the IDH-WT TME program. Taken together, this study provided novel insights into the cellular and molecular features of the TME across glioma molecular subtypes and implied TME as a promising therapeutic target for developing novel treatment strategies.

Lymph node metastasis (LNM) is a major determinant of staging and outcome in cholangiocarcinoma (CCA), yet its mechanisms, biomarkers, anatomical heterogeneity, and treatment evidence remain insufficiently integrated. This review develops an LNM-centered framework in which lymphatic invasion, lymphangiogenesis, genomic alterations and metabolic reprogramming, and reciprocal tumor microenvironment remodeling interact to shape CCA LNM by enabling lymphatic entry, dissemination, immune evasion, tumor-cell survival, and nodal colonization. A subtype-aware synthesis distinguishes small-duct-type and large-duct-type intrahepatic CCA, positions the latter as a biological bridge to perihilar and distal CCA, and maps subtype-associated lymphangiogenic, matrix-remodeling, metabolic, stromal, and immune programs. Translationally, serum ANGPTL4 and exosomal TTN-AS1 are prioritized as candidate adjuncts to preoperative radiologic nodal assessment and CA19-9, whereas tissue ITGB6 represents a comparatively well-characterized candidate for postoperative or biopsy-integrated LNM risk assessment. For node-positive CCA requiring systemic therapy, PD-1/PD-L1 blockade plus gemcitabine-cisplatin provides the first-line backbone. Additional opportunities include intensified chemoimmunotherapy and antiangiogenic combinations, postoperative and locoregional multimodal treatment, genotype-directed or cellular therapies, and mechanism-informed targeting of chemoresistance, oncogenic and lymphangiogenic signaling, stromal and immune niches, and metabolic adaptation. Epigenetic circuitry and post-transcriptional regulation represent emerging regulatory layers, while microbiome-immune interactions provide testable hypotheses for LNM. We propose spatial multi-omics across paired primary tumors, pathologically negative tumor-draining nodes, and metastatic nodes, integrated with pathology-validated liquid biopsy, to identify where metastatic competence emerges and how nodal niches are conditioned. Together, this mechanism-to-clinic synthesis provides a subtype-aware roadmap for biomarker validation, LNM-specific endpoints, and rational therapeutic intervention.

An aberrant mechanical microenvironment is a major driver of malignant tumor progression. Its effects are not limited to the immediate regulation of tumor cell behavior. Through mechanotransduction-mediated epigenetic remodeling, external physical cues can also be converted into persistent malignant programs. Mechanical alterations in the physical tumor microenvironment, including increased matrix stiffness, compressive stress, fluid shear stress, and spatial confinement, can be sensed by integrin adhesion complexes, mechanosensitive ion channels, and the nucleo-cytoskeletal system. These signals are then transmitted to the nucleus through multiple pathways, leading to systematic remodeling of multilayered epigenetic networks, including DNA methylation, histone modifications, chromatin accessibility, and RNA modifications. Notably, some mechanically induced nuclear alterations can be retained even after the original stimulus is attenuated, thereby forming "mechanical memory." This enables tumor cells to maintain malignant phenotypes such as invasion, metastasis, stem-like states, and therapeutic resistance during transitions between distinct microenvironments. In this review, we systematically summarize the key mechanisms by which mechanical signals regulate epigenetic processes across multiple epigenetic layers, with a particular focus on the basis and functional significance of mechanical memory. We also discuss therapeutic strategies targeting mechano-epigenetic coupling nodes and highlight recent advances in clinical translation. A deeper understanding of how mechanical cues are epigenetically written into and maintained as stable malignant states will provide new theoretical insights and potential therapeutic targets for the precision treatment of solid tumors.

Standard chemotherapy effectively induces complete remission in most children with acute myeloid leukemia (AML), however relapse remains the main cause of treatment failure. Leukemia stem cells (LSCs) are proposed to constitute a therapy-resistant reservoir driving disease recurrence, but their reliable identification remains unresolved. We explored intracellular reactive oxygen species (ROS) levels in pediatric AML at diagnosis and sorted ROS-low (RL) and ROS-high (RH) cell fractions. We found that RL cells exhibited increased quiescence, sustained serial colony-forming capacity and engraftment in mice when compared to the RH fraction. Moreover, in a 3D niche model, RL cells preferentially localized to hypoxic regions and were resistant to cytarabine, altogether indicating enrichment for functional LSCs. Transcriptomic analyses revealed enrichment of stemness, adhesion and niche-interaction programs in RL cells, alongside downregulation of mitochondrial and oxidative phosphorylation pathways. Consistently, RL cells displayed reduced oxygen consumption and ATP production, together with remodeled cristae structure despite preserved mitochondrial mass. Paradoxically, complex I inhibition selectively impaired RL cell viability, indicating a dependency on oxidative phosphorylation. Using RL transcriptome, we derived a gene signature that, when applied to single-cell RNA-sequencing data of pediatric AML cases, identified LSCs within the bulk regardless of differentiation status, and showed significant prognostic value. Finally, transcriptional program trajectory indicated that RL fraction contributed to generate the whole AML bulk, whereas gene mutations were maintained in RL and RH cells. Overall, this ROS-based strategy represents a new tool to isolate stem cells transcending traditional markers, and identifies novel clinically relevant vulnerabilities to target LSCs.

GlycoRNAs, a recently discovered category of covalently glycosylated RNA molecules that are predominantly enriched in small non-coding RNAs (sncRNAs) and reportedly displayed on the cell surface and within extracellular vesicles (EVs), have emerged as a novel signaling layer in cancer biology. Through specific interactions with sialic acid-binding immunoglobulin-like lectins (Siglecs), P-selectin, and heparan sulfate-vascular endothelial growth factor (VEGF) complexes, glycoRNAs may contribute to intercellular communication, inflammatory responses, and angiogenesis, highlighting their potential involvement in tumor progression and immune evasion. However, as glycoRNA research remains at an early stage, the clinical translation of glycoRNA-based findings is hindered by a series of fundamental unresolved questions. For instance, the biogenesis and trafficking mechanisms of glycoRNAs are still poorly defined, nucleotide-resolution maps of endogenous glycosylation sites are lacking, detection platforms have yet to be standardized and pervasive platform-dependent quantification bias has not been systematically addressed. In this review, we integrate current advances across the structural, mechanistic, and translational dimensions of cancer-associated glycoRNA biology, and propose a roadmap to advance the field from a fundamental biochemical curiosity toward a clinically actionable target in oncology.

Single-pathway targeted therapy is often insufficient to control malignant glioma progression, and the blood-brain barrier (BBB) remains a major obstacle to effective drug delivery. Therefore, discovering multifunctional target molecules and corresponding drugs that can penetrate the BBB can provide new strategies for targeted glioma therapy. Analysis of public databases and 179 clinical samples showed that solute carrier family 35 member F6 (SLC35F6) was upregulated in gliomas and associated with poorer clinical outcomes. Proteomic analyses and functional validation identified PIK3R2 as a potential downstream mediator of SLC35F6. Mechanistically, SLC35F6 knockdown reduced the abundance of the p85β-p110β complex and promoted proteasome-dependent degradation of p85β (encoded by PIK3R2). These findings support a model in which SLC35F6 maintains p85β stability, at least in part, by preserving the p85β-p110β association, with PIK3R2-associated PI3K/AKT signaling contributing to aerobic glycolysis, glioma cell proliferation, and reduced sensitivity to temozolomide (TMZ). In parallel, SLC35F6 promoted macrophage polarization toward the M2-like phenotype, with evidence supporting the involvement of the macrophage migration inhibitory factor (MIF)/CD74 axis, thereby contributing to an immunosuppressive tumor microenvironment. Leveraging these mechanisms, we developed a dual-ligand nanoplatform designed to facilitate trans-BBB delivery of siSLC35F6, TMZ, and an anti-programmed death-ligand 1 (anti-PD-L1) monoclonal antibody (α-PD-L1). SLC35F6 knockdown suppressed glycolytic reprogramming and induced changes in macrophage polarization consistent with a less immunosuppressive immune microenvironment, thereby enhancing glioma sensitivity to both TMZ and α-PD-L1. Collectively, this study identifies SLC35F6 as a potential therapeutic target for glioma and provides a rationale for further preclinical development of BBB-crossing combinatorial therapeutic approaches.

CD19 CAR-T has achieved high remission rates in relapsed/refractory B-cell acute lymphoblastic leukemia (B-ALL). However, rapid generation with durable in vivo persistence remains a major challenge. CD45RA-negative T cells exhibit sustained antileukemic activity and preserved immune memory. Here, we developed a rapid protocol (InstanCAR-T) enriched for CD45RA-negative T cells, and systematically compared it with conventional CAR-T across in vitro functional assays, murine B-ALL xenograft models, and a Phase I clinical trial. InstanCAR-T shortened ex vivo culture to 3 days while enriching CD45RA-negative subsets, demonstrating enhanced viability, proliferation, and cytotoxicity in vitro. In vivo, InstanCAR-T achieved durable tumor control with improved expansion and long-term persistence. Clinically, high-dimensional immune profiling revealed sustained stemness, reduced exhaustion, and stable phenotypic differentiation of InstanCAR-T cells. Moreover, InstanCAR-T was well tolerated and achieved comparable overall response rates, with a numerically higher proportion of minimal residual disease-negative remission compared with ConvenCAR-T. In conclusion, InstanCAR-T enriched for CD45RA-negative T cells represents a rapid and effective CAR-T strategy with sustained proliferative capacity and superior in vivo persistence. Integrated high-dimensional profiling further highlights its enhanced immunological fitness, supporting its potential to improve clinical outcomes in B-ALL. TRIAL REGISTRATION: This trial was registered at www.clinicaltrials.gov as (NCT05309213) and (NCT06209671).

Platinum-based chemotherapy resistance severely limits therapeutic efficacy and survival in pancreatic ductal adenocarcinoma (PDAC), yet the role of cancer-associated fibroblasts (CAFs) in driving this resistance remains elusive. We screened and validated circMTHFD1L (hsa_circ_0078269) as a stromal circRNA that is markedly upregulated in CAFs from oxaliplatin-resistance (OXA-R) PDAC, and its expression was correlated with poor patient survival. Gain- and loss-of-function assays demonstrated that circMTHFD1L in CAFs is necessary to confer oxaliplatin resistance, primarily through metabolic reprogramming of fatty acid (FA) biosynthesis. Mechanistically, circMTHFD1L directly interacts with the key FA synthesis enzyme acetyl-CoA carboxylase 1 (ACC1), thereby inhibiting its ubiquitination-mediated degradation and enhancing FA production. Tumor cells uptake FA from circMTHFD1L-high CAFs, resulting in increased intracellular uridine diphosphate N-acetylglucosamine availability, a key substrate for O-GlcNAcylation. This metabolic reprogramming promotes Ku70 O-GlcNAcylation and chromatin recruitment, thereby enhancing non-homologous end joining repair efficiency and driving resistance to oxaliplatin in PDAC. Therapeutic targeting of the circMTHFD1L/ACC1 axis in vivo resulted in marked suppression of tumor growth and restored sensitivity to oxaliplatin in orthotopic models. Additionally, we engineered a lipid nanoparticle-based delivery system for sh-circMTHFD1L plasmids and demonstrated that it enhanced the therapeutic efficacy of oxaliplatin in a patient-derived xenograft model. Clinically, high circMTHFD1L expression was positively correlated with ACC1 abundance and may served as a poor prognostic biomarker. Collectively, our study reveals that CAF-derived circMTHFD1L drives oxaliplatin resistance in PDAC by stabilizing ACC1 to reprogram FA metabolism and promote Ku70 O-GlcNAcylation, identifying the circMTHFD1L/ACC1/Ku70 axis as a promising target to overcome chemoresistance.

Prostate cancer is an immunologically cold tumor with a limited response to immune checkpoint blockade. Although tertiary lymphoid structures (TLSs) are associated with favorable immunotherapeutic outcomes in various malignancies, the tumor-intrinsic mechanisms governing their formation in prostate cancer remain poorly defined. We investigated whether the frequently amplified chromatin effector PYGO2 regulates lymphoid organization and antitumor immunity. Using prostate-specific Pten/Smad4 double-knockout and Pten/Smad4/Pygo2 triple-knockout mice, syngeneic tumors, transcriptomics, single-cell RNA sequencing, and immunostaining, we found that Pygo2 ablation restored androgen receptor signaling, reduced neuroendocrine-associated features, delayed castration-resistant progression, and prolonged survival. Pygo2-deficient tumors expressed higher levels of tertiary lymphoid structure (TLS)-associated chemokines, exhibited greater T- and B-cell infiltration, and developed TLS-like structures. Single-cell profiling showed increased CD8/Treg ratios, B-cell abundance, TLS imprint scores, and antigen-presentation signatures, particularly after combination therapy. Pygo2 loss also increased DNA damage and was accompanied by activation of the STING-interferon axis. In RM1 syngeneic tumors, STING silencing attenuated the Pygo2-loss-associated increase in cytokine expression, immune infiltration, apoptosis, and tumor control. Combining Pygo2 ablation with castration, CXCR1/2 inhibition, and dual PD-1/CTLA-4 blockade produced durable disease control in more than 80% of Pygo2-deficient mice. These findings identify PYGO2 as a tumor-intrinsic regulator of castration resistance and antitumor immunity and support combining PYGO2-directed strategies with androgen deprivation, myeloid suppression, and immune checkpoint blockade in prostate cancer.