Advanced Science

Regulatory T cells (Tregs) are key mediators of immune resolution and tissue repair following acute kidney injury (AKI), yet strategies to selectively enrich Tregs within the renal microenvironment remain limited. Interleukin-2 (IL-2) is essential for Treg survival and function, but systemic IL-2 delivery is constrained by a short half-life, narrow therapeutic window, and risk of off-target activation of cytotoxic T cells. Effective Treg-based immunomodulation may therefore require both localized cytokine signaling and targeted recruitment of precursor immune populations. Here, we present a sequential, biomaterial-based immunomodulatory strategy that integrates immune cell recruitment with sustained cytokine delivery. An injectable granular hydrogel physically encapsulating the chemokine CCL21 was used to initially recruit CD4+ T cells to the injured kidney. Subsequently, supramolecular guest-host interactions were leveraged to deliver IL-2 modified with the hydrophobic "guest" moiety adamantane, enabling prolonged local cytokine release over 28 days. In healthy mice, we observed early CD4+ T cell recruitment, followed by a pronounced increase in Treg frequency, consistent with immune reprogramming driven by our sequential delivery system. In a murine model of bilateral ischemia-reperfusion (BiAKI), this injectable biomaterial induced a local increase in the proportion of FOXP3-expressing Tregs and a concomitant decrease in the levels of AKI markers. This two-step immunomodulatory approach minimizes acute immune activation while promoting long-term enrichment of Tregs within the renal microenvironment. Together, these findings demonstrate that granular hydrogels with supramolecular host capability can orchestrate immune recruitment and differentiation in a spatiotemporally controlled manner, providing a versatile immunomodulatory platform for applicability in kidney disease.

Clinical studies have revealed that glioblastoma (GBM) is refractory to immunotherapy, such as immune checkpoint inhibitors (ICIs), due to the immunosuppressive tumor microenvironment (TME) primarily intensified by intratumoral upregulation of transforming growth factor-{beta} (TGF{beta}). Primary GBM tumor bulks are surgically removed in clinic, but invasive tumor cells infiltrated into surrounding normal brain tissue cannot be eradicated by systemic therapy due to the integral blood-brain barrier (BBB), leading to near-inevitable tumor recurrence. To address these translational challenges, we have engineered a TGF{beta}-inhibiting protein nanocage comprising human heavy-chain ferritin protein and TGF{beta} receptor II ectodomain that enables BBB penetration and multivalent TGF{beta} entrapment, respectively. We demonstrate that the nanocage platform, denoted T{beta}R2-HFt, following systemic administration, traverses the BBB, and extensively modulates immunosuppressive TME, both stromal and immune components, thereby potentiating the therapeutic efficacy of a clinically used ICI in multiple mouse glioma models. Notably, a majority of the glioma-bearing animals that received surgical tumor resection, followed by treatment with T{beta}R2-HFt and the ICI, survived the tumor and resisted tumor rechallenge, providing a potential means to control GBM in a durable manner.

Reconstituting organ development in vitro requires reproducing dynamic developmental transitions rather than only terminal tissue structures. Here, we developed MoSAIC (Modular Spatial Assembly and Integration of Cues), a modular platform that decomposes complex culture operations and subsequently integrates them to reconstruct multimodal mechano-chemical spatial cues. Using human induced pluripotent stem cells, MoSAIC enabled independent control of tissue geometry, extracellular matrix localization, physical boundaries, and morphogen gradients, while also permitting defined mechanical and chemical perturbations during development. Their integration guided a continuous ectodermal sheet through neural induction, tissue bending, and bifurcation into neural and epidermal tissues, followed by formation of a polarized neural tube structure. Spatially controlled morphogen exposure further established dorsoventral patterning, while the resulting tissue exhibited posterior neural tube identity. Using the developed model recapitulating the transition of human neurulation, MoSAIC further enables non-genetic perturbation of mechanical and chemical cues to interrogate the mechanisms governing neural tube closure.

The tumor microenvironment in colorectal cancer (CRC) is a heterogeneous ecosystem in which host cells and microbial communities interact dynamically, influencing disease progression. However, the clinical utility is limited by the lack of a scalable spatial host-microbiome technique and by insufficient integration of artificial intelligence for the interpretation of high-dimensional multi-omics data. To overcome the barriers, we present AlphaFISH, a platform technology integrating both technical and computational innovations for multi-omics spatial analysis of clinical biopsies at subcellular resolution. The system uses a sequencing-free, high-throughput, spatial profiling technique to construct, to date, the largest clinical spatial transcriptomics and spatial microbiome datasets acquired from 149 colorectal biopsies from 68 human subjects, supported by a comprehensive scRNA-seq atlas covering 4.27 million cells across 650 patients for robust cell annotation. Deep learning of the cohort-scale dual-omics data, consisting of more than 10 million subcellular sampling vectors, enables the development of a transformer model with joint embeddings of gene expression, spatial architecture, and the microbial microenvironment in colon tissues, achieving nearly 90% accuracy in predicting CRC-associated pathological features using unseen spatial omics inputs. The AI interrogation further predicts tumour recurrence at 81% accuracy in 28 patients followed within 1 year post tumor resection period. AlphaFISH reveals that spatial interactions between Fusobacterium and cellular niche consisting of tumor and T cells serve as key markers of CRC malignancy, progression, and recurrence, indicating the critical role of spatial bacterial-immune crosstalks.

Triple-negative breast cancer (TNBC) is a highly aggressive malignancy and options for targeted therapy are limited. Although PARP inhibitors (PARPi) like olaparib (Ola) leverage synthetic lethality, their clinical utility is severely impeded by drug resistance and low frequency of BRCA mutations. In this study, 1,2,3,4,6-penta-O-galloyl-{beta}-D-glucose (PGG), a natural polyphenolic compound, can induce a state of homologous recombination deficiency (HRD) in TNBC cells by inhibiting the PALB2-BRCA2 interaction. PGG potentiated the inhibitory effects of Ola on the in vitro proliferation, clonogenic potential, migration, and invasion of TNBC cells. Mechanistically, the dual therapy triggered severe oxidative stress and mitochondrial collapse, which accelerated the classical apoptosis cascade and promoted the emission of immunogenic cell death (ICD)-associated DAMPs. The accumulated cytosolic double-stranded DNA activated the innate immune cGAS-STING pathway, thereby promoting robust intra-tumoral infiltration of CD8+ T cells. Furthermore, since cGAS-STING hyperactivation drives compensatory PD-L1 upregulation on TNBC cells, incorporating anti-PD-L1 antibody into the combination regimen further reversed the immunosuppressive barrier, achieving near-complete tumor eradication and maximum infiltration of cytotoxic T cells in vivo. Bioinformatics analyses further suggested a potential network of PGG targets associated with extracellular matrix organization, offering a bioinformatics-derived hypothesis for its theoretical involvement in modulating stromal tension and vascular homeostasis. Collectively, this study establishes PGG as a multi-functional therapeutic agent integrating intracellular synthetic lethality and innate immune sensing, while providing a preliminary bioinformatic rationale for prospective microenvironmental investigation. This study also shows PGG as a multi-functional therapeutic agent that integrates intracellular synthetic lethality, innate immune sensing, and microenvironmental remodeling to convert immunologically "cold" TNBC into "hot" tumors.

Engineered cardiac and skeletal muscle tissues suspended between flexible posts support disease modeling and pharmacology, yet their stimulation, longitudinal imaging and quantitative analysis often remain fragmented and labor-intensive. Here we present Stimulatrix, an open-source platform integrating automated video acquisition within a cell culture incubator, synchronized electrical stimulation and deep learning for longitudinal assessment of contractile tissues in multiwell plates. The platform quantifies tissue compaction, force generation and contraction kinetics with optional cloud processing reducing dependence on local GPU hardware. We demonstrate the workflow in cardiac tissues comprising human induced pluripotent stem cell-derived cardiomyocytes and cardiac fibroblasts, and in primary human skeletal muscle constructs. Force and kinetic measurements were benchmarked against manual annotations. Longitudinal profiling resolved responses to matrix composition and pacing regimens and tracked doxorubicin-associated loss of cardiac contractile force. Stimulatrix provides an accessible workflow for automated functional phenotyping of engineered muscle tissues.

Therapeutic strategies based on increasing uncoupled respiration (browning) in white adipose tissue (WAT) have great potential for treating obesity-related disorders such as diabetes and steatotic liver disease, but have not had a clinical impact because safe and efficient methods for browning WAT have not been developed. Lipid nanoparticles (LNPs) offer a promising platform for adipose-directed siRNA delivery; however, conventional formulations often exhibit limited activity in adipocytes and can induce inflammatory responses that may restrict repeated administration. Here we demonstrate that LNPs that contain ionizable lipids with three lipid tails and siRNAs targeting TLE3 and ZFP423 can efficiently transfect mature adipocytes and induce browning of WAT in vitro and in vivo while providing the biocompatibility needed for repeated subcutaneous injections. Repeated local administration of BLNPs in diet-induced obese mice resulted in sustained target gene silencing, increased expression of thermogenic markers, altered energy metabolism, and decreased hepatic lipid accumulation despite minimal effects on body weight relative to scramble siRNA controls. Collectively, these findings establish BLNPs as a promising platform for adipose-directed RNA delivery and targeted modulation of thermogenic pathways, providing a strategy for investigating how localized adipose tissue remodeling influences metabolic physiology.

Mechanical remodeling by cancer-associated fibroblasts (CAFs) stiffens tumors and can restrict molecular transport, yet the time-resolved coupling between cellular force and matrix mechanical state is poorly defined. Here we use a microfabricated force sensor integrated with 3D CAF-collagen tissue to measure CAF-generated force and collagen stiffness continuously during pharmacological perturbation. ROCK or non-muscle myosin II inhibition reduced force and remarkably drove collagen stiffness below its initial value rather than merely arresting stiffening. This softening was prevented by broad-spectrum MMP inhibition, but only modestly altered by lysyl oxidase inhibition, supporting a competition between force-driven collagen compaction and proteolytic remodeling. Second-harmonic generation imaging showed shorter, thinner and less aligned fibers after ROCK inhibition, while fluorescence recovery after photobleaching indicated faster dextran transport. Moreover, CAF contractile adaptation depended strongly on the duration of inhibition, with reversible recovery after brief treatment, a hyper-contractile rebound following prolonged treatment and washout, and partial force recovery during continuous exposure. These findings identify CAF-matrix mechanics as a dynamic, treatment-history-dependent balance and provide a framework for designing stromal interventions that alter stiffness and transport.

Selectively perturbing pathogenic membranes while preserving host-cell integrity is a fundamental challenge in the design of membrane-active biomaterials. This challenge is particularly acute for enveloped viruses because their lipid membranes are derived from host cells and therefore share many of the same molecular constituents. Here, we demonstrate that engineered nanochirality provides a structural parameter for distinguishing virus-organized membrane nanodomains from host-cell membranes. Histidine-functionalized graphene quantum dots (His-GQDs) form enantiomeric nanoscale structures with distinct interfacial topologies despite closely matched size, composition, and surface charge. D-His-GQDs preferentially interact with cholesterol-rich ordered membrane nanodomains associated with S-acylated coronavirus Spike, penetrate more deeply into raft-like lipid bilayers, and induce lipid disordering, membrane leakage, and viral-envelope disruption. Reducing Spike S-acylation or depleting membrane cholesterol attenuated this stereoselective interaction, while molecular simulations revealed asymmetric lipid organization surrounding S-acylated Spike and preferential insertion of D-His-GQDs into the ordered membrane environment. Functionally, D-His-GQDs directly inactivated human coronavirus OC43 (HCoV-OC43), inhibited infection with an EC50 of 0.92 g/mL and exhibited a selectivity index of 361. Stereoselective antiviral activity extended to human cytomegalovirus, demonstrating activity across distinct enveloped-virus families, and intranasal D-His-GQDs protected HCoV-OC43-challenged mice under both prophylactic and early post-exposure regimens. Overall, these findings establish structural nanochirality as a biomaterial design parameter for recognizing higher-order membrane organization and converting virus-associated membrane states into selectively addressable therapeutic vulnerabilities.

Centromeric heterochromatin is maintained by a self-reinforcing loop between DNA methylation (5mC) and histone H3 lysine 9 trimethylation (H3K9me3), yet whether this coupling is preserved at the level of individual chromosomes in cancer has been inaccessible to conventional epigenomic assays, which average signals across cell populations and cannot resolve single-chromosome heterogeneity. Here, we use antibody-functionalized atomic force spectroscopy to simultaneously map 5mC and H3K9me3 at the centromeres of individual metaphase chromosomes from non-tumorigenic, luminal, and triple-negative breast cancer (TNBC) cell models. Co-localized 5mC/H3K9me3 sites are mechanically stiffer than sites carrying either mark alone across the panel, revealing a cooperative biophysical signature of epigenetic crosstalk. Luminal chromosomes acquire the highest mark density, co-localization, and stiffness, whereas TNBC chromosomes show reduced density and diminished, though not absent, co-localization and stiffness, converging toward the low baseline of non-tumorigenic epithelium. TGF-{beta}1-induced epithelial-mesenchymal transition reproduces this attenuated phenotype in luminal cells, linking it directly to a process central to TNBC aggressiveness. These biophysical parameters classify cancer versus non-tumorigenic chromosomes with near-perfect, cross-validated accuracy, and further resolve the two TNBC subtypes using colocalization-derived features alone (macro-F1 = 0.933). These results establish the mechanical consequence of epigenetic mark co-localization as a single-chromosome-resolvable signature of centromeric organization, attenuated in aggressive breast cancer, and demonstrate a force-spectroscopy platform for mechano-epigenetic profiling of cancer chromosomes.

Experimental surgery is one of the most fundamental fields of biomedical research, recreating clinical conditions on living organisms similar to clinical patients, to facilitate the design of novel therapies. For modeling neurological and psychiatric disorders, the majority of operative techniques mandatorily include craniotomy for consecutive interactions with the open brain. The most wide-spread methodology involves stereotactic surgery in rodents, implementing a specific head holder, attached to the coordinate table, providing the precise animal head positioning according to the brain structures topology. Despite the high efficiency, such systems require complex instrumental skills and long procedure duration per animal even for simple implantation protocols. Hence, novel alternative model species could be a valuable supplementary tool for such manipulations. Therefore, the zebrafish (Danio rerio) fits perfectly, due to its low cost, easy husbandry, rapid development, optic transparency, and high genetic and physiological similarity to rodents and humans. Furthermore, the everted zebrafish telencephalon provides excellent accessibility to critical mammalian-like structures involved in emotion processing and cognition. However, conventional avian- or rodent-inspired craniotomies in adult zebrafish often trigger inadvertent disruption of the brain vessels network, resulting in high operative mortality. A novel, geometry-driven, non-lethal cranial window protocol preserving neurovascular integrity and ensuring 90-100% post-operative survival is presented to overcome above mentioned limitations. This rapid, simple, and high reproducible technique opens new avenues for precise region-specific in vivo neurosurgery, high-throughput psychopharmacological screening, and neurostimulation in a powerful non-mammalian model species

Functional protein micro/nanofibers integrate high specific surface areas with bioactive architectures but remain hampered low production throughput and severe processing instability. Focused rotary jet spinning (FRJS) shows promising to break these throughput constraints while enabling direct, conformal deposition onto complex, irregular substrates. However, navigating FRJS's narrow processing windows in proteins remains failure-prone without closed-loop experimental guidance. Here, we report FiberPro 1.0, a large language model driven multi agent framework that unites high-throughput spinning with real-time experimental feedback for autonomous design, execution, and optimization. Across three protein systems, FiberPro 1.0 converged on spinnable formulations within an average of two iterations. To demonstrate high-throughput conformal deposition in a translational setting, FiberPro designed a zein-based active packaging system applied directly onto diverse food matrices. The resulting conformal coating combined potent antibacterial activity with real-time freshness monitoring, markedly suppressing Escherichia coli and extending shelf life. This work connects autonomous AI reasoning with high-throughput processing, establishing a verifiable route for scalable manufacturing and conformal coating of functional protein micro/nanofibers.

Engineered human neuromuscular models are limited by a trade-off between biological fidelity and architectural control: self-assembled organoids organize stochastically, while engineered 3D tissues impose anisotropy through passive tension at the edges without internal guidance. Here we encode architectural guidance into human neuromuscular tissues by coupling a three-dimensional, cell-laden biomatrix to a xolographically printed hydrogel substrate featuring sub-millimeter surface topography. We demonstrate that grooved substrates increase human primary myotube alignment and maturation in the overlying 3D tissue, while guiding directional axon extension from human iPSC-derived motor neuron spheroids. We show that combining both cell types along a shared alignment axis establishes functional neuromuscular connectivity, evidenced by acetylcholine receptor clustering, glutamate-evoked contraction and its blockade by d-tubocurarine (curare), and yields innervation coverage of centimeter scale tissues. Overall, this platform combines compliant substrates, mesoscale tissue architectural guidance, and human-relevant cell biology in a single untethered construct for studies of neuromuscular tissue development and function.

Leveraging its non-invasive and convenient nature, urinary cytology offers a way for the early diagnosis and surveillance of bladder cancer (BCa). However, its sensitivity remains substantially inferior to the histopathological gold standard, rendering it inadequate to support precise clinical diagnosis. Here, we developed Lipid-featured Urine Cytology with Intelligent Detection (LUCID), an automated platform integrating single-cell stimulated Raman scattering (SRS) imaging and explainable AI model for accurate BCa detection. Following a newly developed standardized urine-preprocessing protocol, 19,678 urinary exfoliated cells from 26 BCa patients and 54 non-BCa participants were imaged using two-color SRS microscopy to simultaneously visualize lipid and protein distributions. For cell-level analysis, a customized deep-supervised Cellpose segmentation algorithm and a YOLO-based detection model were developed to enable precise single-cell segmentation and intracellular lipid droplet (LD) identification, respectively. Totally 20 quantitative morphological and compositional features per cell were extracted and 10 related principal components were obtained via principal component analysis. These features were subsequently integrated into a multiple-instance learning (MIL)-XGBoosting model, which achieved robust diagnostic performance, with 92.31% sensitivity, 87.04% specificity, and an AUC of 0.940. We found that morphological and compositional features provided complementary diagnostic information, while LD-associated features provided primary discriminative information, supporting intracellular LD phenotypes as potential urinary cellular markers of BCa. More importantly, LUCID demonstrated exceptional accuracy in detecting early-stage and small tumors, and showed higher concordance than FISH in a preliminary paired subset (5/6 vs. 2/6). Together, LUCID establishes a chemical imaging-AI paradigm that transforms molecular urine cytology.

Gas-bubble-enhanced focused ultrasound (FUSGB) is advancing clinically across a myriad of applications, including targeted drug delivery through transient blood-brain barrier opening (BBBO). Its broader translation requires a better understanding of the bioeffects that enable reproducible enhancement of barrier permeability while avoiding vascular injury, underscoring the need to clarify the relationship between acoustic exposure, bubble activity, and the biological response central to treatment optimization. These interactions are difficult to isolate in vivo, and acoustic reflections, geometry, and field distortion in conventional cultureware hinder quantitative in vitro studies. Herein, we present a theoretically and experimentally characterized ultrasound-transparent organ-on-chip device with >99.9% US transparency that enables quantitative in vitro modeling of the FUSGB procedure and study of BBBO. Furthermore, submicron bubbles (sMB, ~800nm) were fabricated and characterized. Acoustic dose-dependent subharmonic and broadband cavitation dose curves (SCD and BCD), as well as temporal responses, were studied. A 3-day accelerated Caco--2 surrogate barrier with a high TEER (~1200xcm2) was used to demonstrate US dose-dependent targeted barrier opening, visualized using an in vitro Evans Blue assay. Together, this study and its findings establish a foundation for quantitative in vitro modeling of FUS BBBO, as well as other US interactions in microphysiological systems, supporting systematic evaluation of procedural regimens, large-parameter-space optimization, acoustic safety window, and exposure conditions for brain drug delivery, and additionally enabling mechanistic insight into FUSGB BBBO.

Digital micromirror device (DMD)-based bioprinting enables high-resolution fabrication of hydrogel-based tissue constructs, but print fidelity and hydrogel mechanics are typically coupled through polymer concentration, light exposure, and crosslinking density. This creates a materials constraint for vascular bioprinting, where constructs must preserve microscale architecture during fabrication while supporting multicellular remodeling, lumen formation, and long-term tissue organization. To meet these requirements, we developed a 4D GelMA-Cad/gelatin hydrogel with time-dependent mechanics that maintains print fidelity while providing a soft, cell-permissive matrix. GelMA-Cad provides a covalently photocrosslinked, bioactive network containing an N-cadherin-mimetic peptide, while uncrosslinked gelatin serves as a transient matrix-modifying component that diffuses from the hydrogel after fabrication, producing further softening without bulk degradation of the GelMA-Cad network. We characterized hydrogel chemistry with NMR, quantified time-dependent changes in stiffness, visualized pore morphology with cryo-SEM, and evaluated the platform in retinal and pulmonary constructs. The 4D hydrogel supports 3D-patterned accuracy, cell viability, inner blood-retinal barrier vasculature, and vascularized lung-alveolar tissue models. Single-cell RNA-sequencing of the pulmonary vascular-alveolar model demonstrated compartment-specific transcriptional responses, including enrichment of lung capillary-associated endothelial features and AT1-like epithelial states. Overall, this platform decouples DMD print fidelity from post-print matrix mechanics for vascularized soft tissue bioprinting and tissue modeling.

Capturing fast hemodynamics in deep organs is essential for understanding microvascular regulation of organ responses to physiological demands and pathological stress. However, non-invasive three-dimensional (3D) imaging of these microscale processes remains challenging due to trade-offs between spatial resolution, imaging speed, and penetration depth. Here, we present fluctuation-based high-order super-resolution acoustic microscope (FLAME), a tracking-free 3D ultrasound imaging technology capable of fast microvascular angiography and flow measurement. Using as few as 30 volumes, FLAME improves 3D ultrasound resolution by 8-fold (~50 m) and shortens data acquisition time by over two orders of magnitude, from tens of seconds to tens of milliseconds, compared with conventional tracking-based super-resolution approaches1-6. This high-throughput capability supports high-fidelity imaging of tissue functions under flexible experimental settings. FLAME achieves an unprecedented 3D super-resolution frame rate of ~40 Hz, capturing transient hemodynamic responses to various physiological and pathological stimuli in different mouse organs, from the microvascular level to the whole-body scale. With open-sourced implementation, FLAME provides an accessible platform for real-time, in-depth hemodynamic monitoring across diverse biomedical applications and clinical translation.

Extracellular microRNAs (miRNAs) contribute to metastasis through communication between tumour cells and their microenvironment, yet their sustained, sequence-selective depletion remains challenging. Here we report Programmable RNA-Intercepting Molecular Eraser (PRIME), a membrane-anchored DNA nanomachine that organizes miRNA recognition and catalytic degradation at the cell surface. A double-tetrahedral DNA scaffold positions a programmable recognition hairpin beside RNase H. Target binding generates an RNA/DNA hybrid for RNase H-mediated cleavage, and product release resets the hairpin for subsequent degradation cycles. Membrane-anchored PRIME depleted matched extracellular miRNAs by 70.6-89.4% across lung, breast and cervical cancer cell models. Multi-target PRIME (M-PRIME) inhibited migration by 71.3-96.0% in Transwell assays across these models. To evaluate its therapeutic potential, we established an orthotopic lung-tumour model and an early-intervention breast-cancer lung-colonization model in mice. Aptamer-directed M-PRIME combined with gemcitabine reduced pulmonary tumour burden by 74.5% in established lung tumours and 93.2% in the lung-colonization model. These effects were consistent with PRIME-mediated extracellular miRNA degradation, with circulating target miRNA levels reduced by 77.5-98.7% relative to gemcitabine alone. Together, these findings establish a programmable material strategy for sustained extracellular RNA depletion and support its potential to enhance chemotherapy in preclinical tumour models.

The cornea is the outermost layer of the eye, most vulnerable to injury including from chemical exposure, either environmental, industrial or chemical weapons. For example, exposure of the eye to sulfur mustard (SM) can cause mustard gas keratopathy, resulting in devastating consequences for vision. Available treatments for corneal mustard keratopathy are either ineffective, or if used for a prolonged time, have deleterious side effects. Here, we aimed to harness the anti-inflammatory and anti-angiogenic peptide angiotensin-(1-7), known as Ang-(1-7), that is derived from the renin-angiotensin system (RAS). We designed and developed a protein-like polymer (PLP) displaying a brush of norleucine modified Ang (1-7) peptides (NLE Ang (1-7) PLP), which is a more potent derivative of Ang-(1-7), for administration as an eye drop. In a mouse model of nitrogen mustard (NM)-induced corneal injury, topical treatment and intraperitoneal (i.p.) injection of NLE Ang (1-7) PLP persistently minimized: (i) corneal haze; (ii) edema; (iii), expression of pro-inflammatory and pro-angiogenic genes; and (iv) neovascularization and conjunctivalization as evidenced after a prolonged treatment and recovery period. Collectively, our observations have demonstrated the efficacy of this novel polymer proteomimetic based eye drop treatment in reducing chemical agent-induced corneal injury as well as indicating its potential application in other ocular surface diseases associated with inflammation and angiogenesis.

Spatial transcriptomics reveals cellular organization within intact tissues, whereas dissociated single-cell RNA sequencing provides broad transcriptomic coverage but loses information about cellular proximity and neighborhood structure. Here, we developed ProxiNet, a spatially supervised framework that learns transcriptomic signatures of pairwise cellular proximity from spatial reference datasets and transfers these relationships to dissociated single-cell transcriptomes. ProxiNet predicted cellular proximity across brain regions and spatial technologies, including zero-shot cross-technology transfer, and gradient-based attribution identified genes and broader transcriptional programs associated with proximity predictions. In spatial datasets with known coordinates, ProxiNet-derived cellular neighborhoods recovered reproducible tissue organization and anatomical structure, providing independent spatial validation of the inferred proximity relationships. Applying the spatially calibrated model to dissociated scRNA-seq revealed heterogeneous cellular neighborhoods with distinct cell-type compositions and candidate communication programs. In an Alzheimer dataset, ProxiNet further identified age-associated remodeling of inferred neighborhoods, including an AD-associated neighborhood at 8 months characterized by distinct astrocyte and neuronal transcriptional states and candidate intercellular communication programs. Together, these results establish pairwise cellular proximity as an interpretable and transferable representation for extending spatially learned tissue organization to dissociated single-cell transcriptomes.