Jing Lin

Conventional ferroptosis inducers often suffer from off‐target iron accumulation and insufficient reactive oxygen species (ROS) amplification in tumors, as well as a lack of imaging guidance, and their efficacy is further counteracted by MutT homolog 1 (MTH1)‐mediated sanitization of oxidative DNA damage, collectively limiting therapeutic precision. Here, to circumvent these barriers, we report an engineered nanoinducer integrating ferric ions and a phosphonate‐containing hemi‐cyanine scaffold for real‐time imaging‐guided, MTH1 inhibition‐augmented ferroptosis. This nanoinducer exhibits tumor‐specific accumulation and acid‐triggered dissociation, releasing ferric ions (ferroptosis inducer) and the MTH1 inhibitor TH588. The release kinetics are visualized in real‐time via fluorescence imaging using a phosphatase‐cleavable hemi‐cyanine probe, enabling dynamic therapy monitoring. Mechanistically, ferric ions generate hydroxyl radicals to drive lipid peroxidation and nucleotide oxidation, inducing ferroptosis and DNA mispairing, while TH588 suppresses MTH1 activity, blocking ROS‐induced DNA damage repair and amplifying ROS cytotoxicity. This synergistic strategy significantly enhances antitumor efficacy, suppresses metastasis, and exhibits minimal off‐target toxicity.

Nanocatalytic tumor therapy utilizes nanomaterials to catalyze chemical reactions, producing cytotoxic species that kill tumor cells; however, the metabolic heterogeneity of tumor cells significantly compromises therapeutic efficacy. Herein, we developed a self-amplifying cascade enzymatic nanoplatform (ReS2@GOx) for nanocatalytic tumor therapy by orchestrating metabolic reprogramming and oxidative stress. This nanoplatform integrates rhenium disulfide (ReS2) nanozyme possessing trimodal enzyme-mimetic activities (catalase (CAT)-like, nicotinamide adenine dinucleotide (NADH) oxidase (NOX)-like, and glutathione oxidase (GSHOx)-like activities) with glucose oxidase (GOx), enabling three interlocked therapeutic actions: (1) external energy-independent singlet oxygen (1O2) generation through ReS2-mediated hydrogen peroxide (H2O2) conversion involving a superoxide anion (•O2-) intermediate; (2) attenuation of metabolic circuits through NADH oxidation and glutathione depletion; and (3) a cyclic cascade reaction driven by GOx-mediated glucose oxidation and H2O2 production, coupled with CAT-mediated H2O2-to-O2 conversion, sustaining glucose oxidation and NADH consumption. Mechanistic studies reveal a noncanonical 1O2 generation pathway involving electron shuttle-mediated •O2– conversion, establishing a redox loop fueled by endogenous substrates (e.g., glucose, NADH). The coordinated enzymatic triad achieves dual tumor suppression, including blocking adenosine triphosphate (ATP) production by inhibiting glycolysis/oxidative phosphorylation (OXPHOS) and inducing lipid peroxidation. This work demonstrates a nanocatalytic tumor therapy strategy that orchestrates tumor metabolism to potentiate oxidative lethality, highlighting the therapeutic potential of metabolic regulation for tumor treatment.

Accurate vessel segmentation is crucial for diagnosing vascular diseases and supporting research in developmental biology. However, existing methods struggle to preserve fine capillaries and the topological continuity of vascular structures. In this study, we aim to develop a novel deep learning architecture that specializes in thin vessel segmentation for improved continuity. We propose starU‐Net, a four‐layer encoder‒decoder framework for vessel segmentation. It integrates three components in a problem‐driven design: (1) an enhanced feature extraction module leveraging a “star operation” to enable multiplicative feature interaction, (2) a shallow network design to minimize spatial degradation, and (3) a multi‐view feature fusion module using dynamic snake convolution to capture continuous tubular structures from multiple orientations. The model was trained and evaluated on three public retinal datasets and a novel, self‐constructed chick embryo dataset. Against CNN‐based, transformer‐based, and foundation‐model baselines, starU‐Net obtained the highest sensitivity, F ₁ , area under the curve, and centreline Dice on all three public retinal datasets, at the cost of slightly lower specificity. The improvement in centreline Dice provides quantitative support for the reduction in thin vessel discontinuity. On the chick embryo dataset, whose vessels are markedly wider, the proposed starU‐Net was competitive but not leading. In summary, starU‐Net is an architecture that combines star operation‐based feature extraction with multi‐view tubular feature modeling to improve vascular segmentation. The proposed framework demonstrates strong performance in both clinical retinal imaging datasets and self‐constructed chick embryo angiogenesis data, serving as a promising computational tool for clinical decision support, biomedical research, and large‐scale vascular phenotyping in biomedical applications.

Biomimetic membranes-camouflaged nanomedicines show promising potential in cancer therapy. Herein, we developed biomimetic hybrid membranes-camouflaged biosynthesized melanin nanoparticles, termed MBM-PM, by co-extruding near-infrared (NIR) light-absorbing melanin nanoparticles naturally enveloped in bacterial outer membranes (MBM) with programmed cell death protein 1 (PD-1)-expressing mammalian cell membrane nanovesicles (PM), for efficient cancer photothermal-immunotherapy. The melanin core within the outer membrane vesicles (OMV) generates a photothermal effect, inducing thermal stress to directly kill cancer cells and triggering immunogenic cell death (ICD), which enhances antitumor immunity. Furthermore, the pathogen-associated molecular patterns (PAMPs) present in the bacterial membrane component of MBMs stimulate a robust antitumor immune response. The PM components not only confer cancer cell-targeting capability but also block the PD-1/programmed death-ligand 1 (PD-L1) interaction, further enhancing immune activation. Our studies demonstrate that the MBM-PM nanoplatform can effectively eradicate primary tumors and significantly inhibit distant tumors and lung metastasis, offering a promising biosynthesized nanoplatform for cancer photothermal-immunotherapy.

Deep learning-based key point detection has been widely applied in medical image analysis. However, these models are usually obtained through supervised training on specific datasets and for specific tasks. Without re-conducting data collection, manual annotation, and re-training, such supervised models are generally difficult to adapt to users’ new demands in various clinical scenarios. To address this challenge, we first synthesize a large-scale key point detection dataset covering different imaging modalities and anatomical regions using public medical image segmentation datasets. To better meet the long-tail needs of clinical users, inspired by the idea of in-context learning, we further develop a user-customizable key point detection model (named KeyPointGPT) based on the in-context segmentation model SegGPT, which allows customized key point detection with only a few visual examples. Users only need to provide a small number of key point examples (e.g., 1-3 cases), and the pre-trained KeyPointGPT model can automatically perform subsequent key point detection according to users’ requirements, greatly improving the flexibility and applicability of the proposed key point detection model. Extensive experiments demonstrate the effectiveness of our model on unseen datasets and tasks. By improving flexibility through in-context learning trained on large-scale synthetic data, the KeyPointGPT model is expected to address the currently unmet long-tail demands of customized key point detection in clinical applications.

Although nitric oxide (NO) holds great therapeutic potential in oncology, achieving its precise spatiotemporal release and understanding its synergistic molecular actions remain as major challenges. Here, we developed an acid/near-infrared (NIR) dual-responsive nanoplatform (namely, BCN) with AND logic-gated controlled NO release for companion theranostics (CTx) of tumors. BCN is assembled by an acid-responsive photothermal agent (Croc880), a thermoresponsive NO donor (BNN6), and human serum albumin. Specifically, the acidic tumor microenvironment can activate the photothermal effect of Croc880, which subsequently triggers the decomposition of BNN6 and releases the NO. Meanwhile, the acid-responsive fluorescence (FL) and photoacoustic (PA) properties of Croc880 provide real-time NIR-II FL/PA imaging readouts, enabling visualization of the activation process and spatiotemporal guidance of laser irradiation, thereby facilitating imaging-guided CTx. Notably, the released NO disrupts the tumor thermal defense system by inhibiting autophagy and heat shock protein 70 expression. This reciprocal reinforcement between NO signaling and mild hyperthermia results in a strong synergistic effect that effectively eradicates tumors, offering a versatile strategy for the CTx of tumors.

Immune checkpoint blockade (ICB) often fails against immunologically “cold” tumors, a challenge exacerbated by their reprogrammed metabolism. The Warburg effect drives tumor cells to rely on aerobic glycolysis, resulting in excessive lactate production. While simultaneous deprivation of glucose and lactate using glucose oxidase (GOx) and lactate oxidase (LOx) holds promise for starving tumors and alleviating immunosuppression, the activity of both enzymes is greatly constrained by tumor hypoxia. To overcome this, we develop a biomimetic nanoemulsion (FGL@PM) coloaded with GOx, LOx, and perfluorotripropylamine (FTPA), and coated with a programmed cell death 1 (PD-1)-presenting cell membrane. FTPA, owing to its high affinity for oxygen, can carry and deliver a large amount of oxygen for supporting the catalytic reactions of both GOx and LOx. This nanoemulsion enhances ICB via glucose and lactate cooperative deprivation (GLCD) strategy. In detail, surface PD-1 proteins not only blocked PD-ligand 1 (PD-L1) on tumor cells but also promoted the cellular uptake of FGL@PM. GOx catalyzed glucose oxidation, generating H₂O₂ and suppressing glycolysis and lactate production, while LOx further depletes lactate and boosted H₂O₂ generation. The resulting H₂O₂ disrupts the redox homeostasis in tumor cells and stimulates antitumor immunity. This GLCD strategy effectively suppressed tumor growth on both unilateral and bilateral subcutaneous 4T1 breast tumor models by simultaneously impairing glycolysis and oxidative phosphorylation, enhancing antigen presentation, and promoting the infiltration of antitumor T cells. This study highlights the potent combination of glucose/lactate metabolic intervention with immunotherapy for improved antitumor treatment.

Triple-negative breast cancer (TNBC) is characterized by a lack of effective therapeutic targets, high invasiveness, and a propensity for metastasis, making the development of highly specific and sensitive early diagnostic and personalized treatment strategies particularly urgent. Leveraging the distinct features of the TNBC tumor microenvironment─specifically, acidic pH and abnormal Cu²⁺ enrichment─this study designed and constructed a pH/Cu²⁺ dual-responsive ratiometric photoacoustic (PA) nanoprobe, termed LPCR NP, for precise in vivo PA imaging and the differential diagnosis of TNBC. The nanoprobe consists of a pH/Cu²⁺ dual-responsive cyanine dye (LET-PCu) and a reference dye (IR140), exhibiting excellent biocompatibility and stability. In vitro experiments confirmed that under acidic conditions and in the presence of an elevated Cu²⁺ concentration, the probe exhibited a significant increase in optical absorption, which led to a robust enhancement of PA signal at 720 nm, demonstrating excellent dual-activation behavior. Meanwhile, IR140 maintained a stable reference PA signal at 820 nm, allowing the LPCR NP to generate highly distinguishable ratiometric PA signals. In a subsequent in vivo double-blind study, the nanoprobe successfully differentiated between MDA-MB-231 (TNBC) and MCF-7 (non-TNBC) tumor models through PA imaging. This nanoprobe not only provides a novel tool for noninvasive and accurate diagnosis of TNBC but also holds significant translational potential in theranostic applications such as monitoring the efficacy of copper depletion therapy and screening eligible patients.