Biological groups exhibit an intrinsic capacity for self-organization, engendering complex and intelligent cooperative behaviors—a capability highly sought after for robotic swarms. This paper proposes an architecture for constructing swarm intelligence in cooperative tasks, termed Cooperation Emergence and Strategy Generation (CE&SG). The architecture empowers robot swarms with the ability to dynamically synthesize composite strategies through self-organized interactions, setting it apart from traditional passive or fixed approaches. CE&SG is inspired by biological observations of cooperation in primate groups, from which we extract an explanatory mechanism for self-organized strategy generation and implement it in robot swarms using a hyper-heuristic algorithm. Through this approach, we demonstrate that sophisticated swarm cooperation can emerge via continuous strategy generation. Concurrently, the framework addresses long-standing challenges in robotics, including adaptability to dynamic changes, system scalability, and strategic interpretability. In a proof-of-concept task of air defense suppression, the proposed approach outperforms several existing advanced methods. The task is executed collaboratively by robots with diverse functionalities. We tested scalability and fault tolerance mechanisms against multiple disruptive disturbances at scales of up to 250 robots and 1,500 tasks in scheduling-oriented discrete-event simulations. Physical feasibility is further validated in a high-fidelity environment with continuous kinematics and collision constraints. The evolutionary patterns of strategies are analyzed through case studies.

Genetically engineered probiotic bacteria are promising living therapeutics. However, their therapeutic potential is frequently confined to their colonization sites, and limitations in efficacy, controllability, and robustness remain major barriers to translation. Anti-inflammatory cytokines play a crucial role in mitigating the pathogenesis of inflammation-related diseases, yet their application is limited by their vulnerability in the mammalian gastrointestinal tract. In this regard, we first found that interleukin-37b (IL-37b) mitigated inflammatory responses at picogram concentrations in primary intestinal cells, supporting its translational potential for the oral treatment of inflammatory diseases. Then, we harnessed the probiotic bacterium Escherichia coli Nissle 1917 as a chassis and further engineered it into a hypervesiculating strain, ΔECIL-37b, capable of efficiently secreting IL-37b-bearing bacterial outer membrane vesicles (OMVs) in situ. These OMVs facilitated IL-37b penetration across the intestinal barrier, enabling it to reach the inflamed mucosa and interact with target cells. In both acute and chronic murine colitis models, we demonstrated that ΔECIL-37b achieved anti-inflammatory efficacy comparable to direct IL-37b injection while enabling local intestinal delivery. Mechanistically, ΔECIL-37b exerted anti-inflammatory effects by attenuating myeloid differentiation primary response 88 (MyD88)-related downstream extracellular-signal-regulated kinase/nuclear factor κB signaling and inflammatory-immune cross-talk. Together, these findings provide a proof-of-concept strategy to harness the anti-inflammatory activity of IL-37b via probiotic-derived OMV delivery and offer insights into a precision anti-inflammatory approach integrating cytokine and probiotic functions.

While critical for human–machine interfaces, on-skin touch sensors struggle to replicate human skin’s curved deformation mechanics and achieve simultaneous high sensitivity, linearity, robustness, and stability. Existing capacitive, piezoresistive, piezoelectric, or triboelectric mechanisms cannot effectively harness touch-induced curved indentation. We introduce a skin-inspired flexoelectret touch sensor that exploits flexoelectricity—converting strain gradients into electricity—to mimic skin’s curved deformation geometry. This enables self-powered, self-calibrated tactile imaging through a novel mechanism establishing linear correlation between inhomogeneous deformation and electrical signal. The sensor achieves high sensitivity (the highest sensitivity up to ~835 mV/N), good linearity, exceptional robustness (>80,000 s continuous operation; >8,000 cycles), and immunity to temperature fluctuations, ultraviolet light illumination, and humidity interference. A 7 × 7-pixel array demonstrates shape-adaptive and touch-trajectory imaging capabilities. This technology unlocks transformative potential for wearable electronics and advanced human–machine interactions.

D-peptides are attractive therapeutic modalities because they are generally more resistant to proteolysis than their L-counterparts, yet systematic design of target-binding D-peptides remains nontrivial. Here, we report Mirror-Peptidizer, an end-to-end in silico mirror-image screening workflow that generates D-peptide binders without requiring chemical synthesis of D-protein targets. The workflow mirrors an L-protein structure to a virtual D-protein, designs L-peptide backbones in the presence of the mirrored target using a diffusion-based backbone generator, selects sequences with a neural sequence design model, and explores local sequence neighborhoods via Bayesian multi-objective optimization balancing sequence-backbone compatibility and a solubility heuristic. Mirroring the resulting complex yields the corresponding D-peptide predicted to bind the native L-target. Using MDM2, PD-L1, and interleukin-23 receptor (IL-23R) as test cases, we identified D-peptides spanning α-helical, β-rich, and mixed conformations with affinity from 11.9 nM to sub-μM. For the MDM2 system, the ¹H-¹⁵N HSQC (heteronuclear single quantum coherence) perturbations and protein mutagenesis support the designed interface, and cell-penetrating conjugates show p53-dependent cancer growth inhibition. Similarly, PD-L1-targeting D-peptides potently inhibited PD-1/PD-L1 interactions in competitive binding assays in vitro, and IL-23R-targeting D-peptides inhibited IL-2/IL-12/IL-23-induced interferon-γ production in human peripheral blood mononuclear cells. Mirror-Peptidizer is provided as an open-source implementation to facilitate rapid generation of experimentally testable D-peptide starting points.

Establishing physiologically relevant in vitro tumor models is critical for accurately evaluating chemotherapeutic efficacy in ovarian cancer. However, generating high-throughput, uniform, and viable tumor spheroids from primary patient cells remains a major challenge for personalized drug screening. Here, we present an integrated platform that couples hydrogel microencapsulation with organ-on-a-chip culture to achieve biomimetic 3-dimensional tumor growth and dynamic drug assessment. Core–shell hydrogel microcapsules, consisting of a carboxymethyl cellulose core and an alginate shell, create a biocompatible and tunable microenvironment that maintains cell viability, preserves molecular and phenotypic heterogeneity, and enables reproducible spheroid formation. Embedding these spheroids into a microfluidic chip with continuous, precisely regulated drug gradients and perfusion culture facilitates high-resolution, high-throughput evaluation of chemotherapeutic responses. Using this system, clinically relevant agents—including carboplatin, paclitaxel, docetaxel, and pegylated liposomal doxorubicin—were systematically screened, revealing patient-specific variations in drug sensitivity that closely aligned with postoperative clinical outcomes. By integrating hydrogel-based 3-dimensional culture with microscale gradient control, this approach provides a stable, physiologically meaningful, and scalable platform for preclinical pharmacological testing. Collectively, the findings demonstrate its potential as an effective tool for individualized chemotherapy evaluation and precision treatment development in ovarian cancer.

Boron neutron capture therapy (BNCT) enables localized tumor ablation while minimizing damage to surrounding tissues, offering advantages for treating anatomically challenging sites. However, current boron carriers, such as sodium borocaptate (¹⁰BSH), suffer from inadequate tumor specificity. Herein, the present study details the design, synthesis, and preclinical evaluation of a novel small-molecule boron-10-enriched carrier, which was synthesized by covalent bond coupling 4-carboxy-3-fluorophenylboronic acid (FPBA) to ¹⁰BSH (FPBA–BSH), achieving a boron content of approximately 25 wt.%. FPBA–BSH efficiently penetrated the blood–brain barrier and demonstrated pronounced accumulation in orthotopic gliomas, achieving a boron concentration of 75.4 μg-B/g-tumor tissue, which was 4.2- and 3.7-fold higher than those with boronophenylalanine (BPA) and BSH, respectively. Moreover, FPBA–BSH exhibited markedly improved tumor selectivity, with tumor-to-normal tissue (T/N) and tumor-to-blood ratios of 52.0 and 7.2, respectively. The T/N ratio was approximately 19.3- and 14.1-fold greater than those observed for BPA and BSH. In the melanoma model, FPBA–BSH achieved an intratumor boron concentration of 114.4 μg-B/g-tumor tissue, representing 8.4- and 9.9-fold increases compared with BPA and BSH, respectively. Correspondingly, the T/N and tumor-to-blood ratios reached 135.1 and 8.6, indicating substantially enhanced tumor targeting and retention. The T/N ratio achieved with FPBA–BSH was approximately 26.0- and 34.6-fold higher than those obtained with BPA and BSH, respectively. Consistent with its superior tumor selectivity, FPBA–BSH-mediated BNCT induced pronounced tumor-selective cytotoxicity and markedly inhibited tumor growth in both orthotopic glioma and melanoma models compared with BPA, BSH, and untreated controls. These findings demonstrate that FPBA–BSH represents a promising small-molecule boron delivery agent with substantial potential for clinical BNCT applications.

The advancement of porous carbon supercapacitors is constrained by the inability of conventional characterization to probe their dynamic electrode–electrolyte interfaces. In-situ characterization bridges this gap, offering real-time observation of critical processes such as ion adsorption/desorption and surface reactions under operating conditions. This Perspective illustrates how these in-situ techniques are transforming mechanistic understanding, discusses persistent challenges including multimodal integration and spatiotemporal resolution trade-offs, and proposes a roadmap to translate these insights into the design of next-generation devices.

Schistosomiasis, a neglected tropical disease, affects more than 250 million people worldwide. The pathology and transmission of schistosomiasis rely on the parasite’s extended survival and massive egg output. While N⁶-methyladenosine (m⁶A) messenger RNA modification is an important epigenetic regulator, its functional roles in pathogenic helminths remain entirely unexplored. Here, we comprehensively characterize the m⁶A landscape in the human blood fluke Schistosoma japonicum. We generated high-resolution transcriptome-wide m⁶A maps for adult male and female parasites, revealing distinct sex-specific methylation patterns. Functional investigation of the conserved m⁶A writer complex demonstrated that m⁶A methyltransferase-like-3/14 (METTL3/14)-mediated methylation is indispensable for female reproductive development and oviposition. In males, m⁶A modification is essential for maintaining tegumental integrity, metabolic homeostasis, and motility. Mechanistically, integrated methylated RNA immunoprecipitation sequencing and RNA sequencing analyses following METTL3 knockdown identified key m⁶A-modified transcripts, including glutamine synthetase, which underpin the observed phenotypic defects. Our study provides the first functional evidence of m⁶A-mediated epitranscriptomic regulation in a parasitic helminth, establishing it as a master regulator of reproductive and somatic development and revealing the m⁶A pathway as a potentially important target for developing novel strategies against schistosomiasis.

Advancements in biomedical interfaces increasingly demand technologies that can seamlessly bridge the gap between biological tissues and therapeutic systems. Microneedle (MN) technology has emerged as a minimally invasive platform for transdermal drug delivery (TDD) and biosensing, offering tunable geometries, efficient skin penetration, and reduced patient discomfort. However, the inherent limitations of conventional microfabrication techniques in terms of structural complexity, multi-material compatibility, and functional modularization have markedly constrained the development of next-generation biomedical systems. In recent years, 3-dimensional (3D) printing has positioned itself as a highly promising additive manufacturing (AM) technology, offering exceptional design freedom and high-resolution fabrication capabilities for the development of MN platforms with embedded microchannels and integrated multifunctionality. Herein, a comprehensive analysis of recent progress in 3D-printed MNs is provided, with emphasis placed on advances in architectural innovations, intelligent system integration, and their expanding applications in personalized drug delivery and intelligent theranostic platforms. Furthermore, an in-depth examination of the core challenges hindering the clinical translation of 3D-printed MNs, particularly regarding manufacturing processes, material selection, and standardization requirements, is presented, offering a forward-looking perspective on the paradigm shift of MNs from passive delivery terminals to active, closed-loop health platforms.

In middle-aged and older atherosclerotic renal artery stenosis (ARAS), the anatomical severity of stenosis is a poor surrogate for microvascular competence, and the renal benefit of revascularization is unpredictable. We developed Renal-Video-AI, a self-supervised deep learning framework (Video Swin Transformer with VideoMAE pretraining) that extracts spatiotemporal hemodynamic features from contrast-enhanced ultrasound, and applied it to a multi-center Discovery Cohort (N = 1,226), an independent External Validation Cohort (N = 122), a prospective Multimodal Cohort with paired 10x Visium spatial transcriptomics (N = 57), and an aged two-kidney-one-clip (2K1C) murine model. Unsupervised phenomapping identified 3 intrinsic hemodynamic phenotypes—Preserved, Delayed, and Rarefied. The Rarefied phenotype predicted major adverse renal events (MAREs) independently of anatomical stenosis [hazard ratio (HR) 4.82, 95% confidence interval (CI) 3.10 to 6.50; Fine–Gray subdistribution HR (sHR) 5.1], and adding the phenotype to a standard clinical model improved the C-statistic from 0.72 to 0.88. A significant phenotype-by-treatment interaction (P < 0.01) showed that stenting reduced events only in the Delayed phenotype (HR 0.52, 95% CI 0.35 to 0.78), not in the Preserved (HR 0.98) or Rarefied (HR 1.05) phenotypes. In absolute terms, stenting reduced the 3-year cumulative incidence of MARE in the Delayed phenotype from 25.4% to 13.2% (absolute risk reduction 12.2%; number needed to treat = 8, 95% CI 6 to 13), with no benefit in the Preserved (8.4% versus 8.0%) or Rarefied (38.6% versus 39.4%) phenotypes. Spatial transcriptomics localized a hypoxia and pyroptosis signature to rarefied tissue, and the aged 2K1C model revealed a mitochondrial reactive oxygen species (ROS)–NLRP3–pyroptosis axis whose pharmacological inhibition (MCC950) restored microvascular perfusion. AI video-phenomapping thus reframes the revascularization decision around microvascular competence rather than anatomy, identifying both therapeutic futility (Rarefied) and a treatable window (Delayed), and nominates NLRP3-driven pyroptosis as a therapeutic target.