Biology subjects

Liu, T.

Publications and source records attributed to Liu, T..

At least 55 records · Page 3Linked to original sources

bioRxiv2026

Antibodies to influenza A virus hemagglutinin and neuraminidase limit egress and alter the physical properties of released virus particles

Influenza A virus (IAV)-specific antibodies neutralize mature virions by inhibiting functional sites or, in some cases, by promoting virion aggregation. Many antibodies also act on infected cells to reduce virion yields, but the underlying mechanisms and effects on the physical properties and function of released virus particles remain incompletely characterized. Here we use flow virometry to acquire high-sensitivity yield and size measurements of virus particles released in the presence of antibodies. Combined with digital-droplet PCR and electron microscopy, this approach enables comprehensive characterization of antibody-induced changes in particle genome-content and morphology. We show that antibodies rapidly and dynamically alter released particle distributions, reducing yields and inducing the production of larger particles. Both effects result in part from aggregation induced by crosslinking of viral antigen on the infected-cell surface and inhibition of viral NA. However, yield reduction is not fully explained by aggregation, and a subset of induced larger particles are elongated virions. Finally, particles formed in the presence of HA stem-binding antibody, which does not inhibit attachment of mature virions, show reduced attachment in subsequent rounds of infection. Altogether, we uncover an unappreciated mechanism by which antibodies interfere with viral infection that occurs only during budding and release. Our work highlights the necessity of studying how the immune response shapes virus populations in the context of active infection processes.

microbiology↗

bioRxiv2026

Monoclonal nephritic factors reveal insights into C3 convertase dynamics and dysregulation

C3- and C5-nephritic factors are potent but poorly understood autoantibodies that dysregulate complement convertases. To date, their underlying mechanisms of action, epitopes, and sequences remain unknown. To address this knowledge gap, we immune profiled B cells from a nephritic factor-positive C3 glomerulopathy patient and identified the first monoclonal C3- and C5-nephritic factors. We present the structure of a C3-nephritic factor bound to a C3 convertase with the convertase protease domain unexpectedly rotated and inhibited. This rotation advances our understanding of complement convertase progression and decay, explains disease-associated convertase variants, and reveals the molecular mechanism by which nephritic factors can either activate or inhibit convertase activity. We also detail heterogeneity within and between C3- and C5-nephritic factors in terms of convertase binding, stabilizing capacity, regulator inhibition, fluid-phase activation, and disease contribution. These findings improve stratification of patients with C3 glomerulopathy, redefine basic C3 convertase dynamics, and provide insights into antibody-mediated modulation of the complement system.

immunology↗

bioRxiv2026

STAG2 Maintains Chromatin Compartmentalization and Represses Regulatory Element Contact to Promote Oncogenic Signaling in Muscle Invasive Bladder Cancer

Contrary to other cancer types, stromal antigen 2 (STAG2) expression is associated with shorter survival and an invasive phenotype in muscle invasive bladder cancer (MIBC). As a cohesin complex component, STAG2 regulates genome organization and cell type-specific transcription, yet its mechanistic role in MIBC remains unclear. Here, we uncover mechanisms through which STAG2 coordinates chromatin architecture and gene regulation in MIBC. Modulation of STAG2 rewired chromatin architecture, altering chromatin contacts and compartmentalization, increasing promoter-enhancer interactions, and reducing short-ranged chromatin loops. At specific gene loci, we discovered that STAG2 has context-specific activating and repressive regulatory functions. At the STAG2-activated gene ABCA1, STAG2 maintained A-compartment chromatin and high levels of promoter acetylation, indicators of active transcription. STAG2 KO resulted in loss of promoter acetylation, a shift from A to B compartment chromatin, and increased occupancy of the co-repressor TRIM28, resulting in ABCA1 downregulation and diminished invasive potential. Conversely, at the STAG2-repressed gene SPOCK3, STAG2 KO resulted in B to A compartment switching, aberrant formation of chromatin loops, and SPOCK3 upregulation. Treatment with EZH2 inhibitor tazemetostat augmented STAG2-KO induced SPOCK3 upregulation, suggesting a collaborative role of STAG2 and EZH2 in repressing Polycomb Repressive Complex 2 (PRC2) target genes. Altogether, our results indicate that STAG2 plays a multifaceted role in regulating gene expression in bladder cancer that is dictated by the epigenetic and chromatin landscape of the cells. These findings identify STAG2-dependent vulnerabilities and provide a rationale for therapeutic targeting of chromatin regulators in MIBC.

Cancer Biology↗

bioRxiv2026

eQTM (expression quantitative trait methylation) Atlas: a comprehensive resource of over 11 million DNA methylation-gene expression associations through across 11 tissues and 4 diseases

MotivationEpigenome-wide association studies (EWAS) have identified numerous DNA methylation (DNAm) CpG sites associated with complex traits and diseases, but interpretation of those CpG sites remains challenging because in EWAS, CpGs are mostly linked to nearby genes based only on genomic proximity. Expression quantitative trait methylation (eQTM) analyses connect DNAm CpGs with statistically associated gene expression levels. However, a comprehensive, searchable resource integrating eQTMs across diverse tissues and disease contexts has been lacking. ResultsWe developed the eQTM Atlas, a web-based resource that manually curates more than 11 million DNAm-gene expression associations from eight cohorts, covering 11 tissue types, four broad disease contexts, 173,886 unique CpG probes and 20,231 unique genes. The Atlas supports gene- or CpG-searches by tissue or disease type and finding associated CpG or genes, visualization of cis- and trans-eQTMs through genome browser, heatmap interfaces across various tissues, and cohort-level data downloads. By integrating eQTM results with EWAS resources, the eQTM Atlas enables users to connect disease- or trait-associated CpGs to statistically associated genes rather than relying solely on proximity-based gene annotation, supporting functional interpretation of EWAS findings and generation of disease-specific regulatory hypotheses. Availability and implementationThe eQTM Atlas is freely available at https://shiny.crc.pitt.edu/eqtm_browser/. The web interface is implemented in R Shiny and hosted through the University of Pittsburgh Center for Research Computing (CRC). Source code is available at https://github.com/ads303/eQTM-Atlas.

genetics↗

bioRxiv2026

An SE(3)-equivariant Crystal Structure Prediction Framework for Prospective Identification and Development of Bioactive Nucleoside Self-Assembling Materials

Flexible organic small molecules can assemble into supramolecular biomaterials whose properties are intrinsically governed by their crystal structures, yet experimental structure determination remains difficult to scale during molecular modification and materials optimization. Crystal structure prediction (CSP) provides a potential solution, but its prospective power is limited for flexible molecules owing to incomplete conformational sampling and the difficulty of identifying experimentally realized structures from large candidate sets. Here, an SE(3)-equivariant deep-learning workflow, SE3CSP, is developed for organic crystal structure prediction. By learning molecular conformations, unit-cell parameters and packing patterns from experimentally resolved crystal structures and integrating these predictions with MACE-based structure optimization, SE3CSP establishes an end-to-end pipeline from two-dimensional molecular representations to three-dimensional crystal structures. Using nucleosides as representative flexible self-assembling building blocks, SE3CSP achieves an overall prediction accuracy of [~]57%, substantially outperforming the benchmark MACE-based Genarris 3.0 workflow ([~]14%). Furthermore, a prospective prediction strategy is developed in which an SE3CSP-predicted density window ({+/-} 0.15 g/cm3) is applied prior to energy ranking and structural deduplication, enabling all experimentally realized structure to be consistently ranked within the top 1-2% of all generated candidates. Beyond structure prediction, SE3CSP-derived energy landscapes provide insight into potential single-crystal-to-single-crystal transformations and enable the identification of a nucleoside supramolecular material with dynamic breathing porosity, which is further developed as an adsorptive platform for inflammatory mediator removal with excellent anti-inflammatory performance and biocompatibility. These results establish SE3CSP as a practical framework for prospective CSP and highlight its utility in guiding the discovery and design of bioactive self-assembled materials.

bioengineering↗

bioRxiv2026

Extending structural surfaceomics to identify aberrant conformations of tumor surface proteins as potential immunotherapy targets

The complement of tumor cell surface proteins, or "surfaceome", is a rich source of potential immunotherapy targets. To move beyond expression-based target discovery, we previously described "structural surfaceomics," combining crosslinking mass spectrometry (XL-MS) with surface protein biotinylation to identify conformation-selective targets. In our prior work, we applied this method to a single model of acute myeloid leukemia (AML), identifying active integrin beta-2 as a promising target. Here, we expand structural surfaceomics to identify additional immunotherapy targets and surface protein biology across additional models of AML, multiple myeloma, and prostate cancer, as well as donor peripheral blood mononuclear cells. Utilizing these models and different chemical crosslinkers, we compile an extensive database of 5,209 crosslinks. We characterize both shared and unique crosslink-based features, identifying 1,612 disease model-specific crosslinks, including 212 potentially defining tumor-specific conformations based on distance constraint violations relative to AlphaFold predictions. We further implement a suite of emerging modeling tools to predict tumor-specific protein structures. We probe crosslinking patterns suggesting multiple myeloma-specific CD48 and AML-specific integrin 1/{beta}4 heterodimer conformations. This work establishes a resource for cancer structural biology by implementation of structural surfaceomics. Our findings also point toward more realistic protein design models, potentially enabling systematic detection of targetable cancer-specific epitopes for next-generation immunotherapies.

cancer biology↗

bioRxiv2026

3D Printed Bioelectronic Scaffolds for Impedance-based Cytotoxicity Monitoring of In Vitro Cancer Models

Cancer is a significant contributor to global mortality and places a substantial burden on healthcare systems, underscoring the need for improved strategies for developing and evaluating new therapies. Electrochemical impedance monitoring of in vitro cancer models is a promising technique for evaluating treatment effectiveness, particularly for evaluating how well a drug may kill cancer cells. This approach is advantageous over conventional end-point assays because it is non-destructive, label-free, and can provide temporal information on cell behavior and drug kinetics. However, traditional impedance devices are limited in that they do not support three-dimensional cell culture that has become standard in cancer studies. Typical devices are planar substrates that support monolayer culture, which has been shown to overestimate drug effectiveness. In this work, we propose 3D printed bioelectronic scaffold devices that provide 3D cancer cell culture while functioning as an on-chip readout for monitoring changes in cell characteristics via impedance. We describe device development and demonstrate reproducible fabrication, stable electrochemical properties, cell detection by impedance, and proof-of-concept monitoring of cytotoxicity in response to a chemotherapeutic drug. Overall, this technology offers a promising platform that could be further developed for compound screening as part of drug development or precision medicine.

bioengineering↗

bioRxiv2026

Additive manufacturing of PEDOT:PSS electrodes on collagen substrates for soft and bioactive electronics

Traditional bioelectronic devices are limited by poor biointerfacing due to their substantial mismatch in mechanical and biochemical properties. In tissue engineering, soft and bioactive materials support biointegration by harnessing or mimicking the natural extracellular matrix (ECM). Building bioelectronic devices from ECM should improve their biointegration, yet there are limited methods to fabricate them due to current manufacturing approaches. An additive manufacturing strategy is presented here for collagen-based bioelectronic interfaces that integrates conducting polymer electrodes with ECM-based substrates or encapsulation layers. Addition of poly(ethylene glycol) diglycidyl ether (PEGDE) to poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate) (PEDOT:PSS) colloidal dispersions enables direct extrusion-based patterning under mild conditions compatible with collagen substrates, and forms aqueous stable and highly conducting printed patterns (2788 S m-{superscript 1}). The resulting interfaces maintain stable electrochemical performance over 7 days in physiological environments, and support primary human cell adhesion, viability, and proliferation across both material regions. A sacrificial patterning strategy using 3D printed cacao butter further enables spatial control of collagen encapsulation. This approach establishes a framework for fabricating functional bioelectronic devices based on ECM to further enhance device biointerfaces for tissue models and implantable systems.

bioengineering↗

bioRxiv2026

Microglial pruning of extinction-ensemble synapses preserves fear memory

Fear extinction suppresses learned fear without erasing the original memory, yet how competing fear and extinction ensembles are selectively updated remains unknown. Here, we show that microglia preserve fear memory by selectively editing extinction-ensemble synapses during retrieval. In medial prefrontal cortex, microglial processes, but not somata, expand, ramify, and tighten their engagement with extinction-ensemble dendrites and spines, where they preferentially engulf excitatory postsynaptic material and bias spine remodeling toward elimination. This selectivity is instructed by local find-me, eat-me, and dont-eat-me cues: purinergic signaling recruits microglial processes, phosphatidylserine exposure licenses engulfment of extinction-ensemble synapses, and CD47-SIRP protects fear ensembles from removal. Weakening microglial recruitment or engulfment, or removing this protection, accelerates extinction without impairing fear acquisition. These findings identify microglial processes as active gatekeepers of ensemble competition and reveal a neuroimmune mechanism that preserves fear memory by limiting extinction. In BriefDuring memory updating, microglial processes selectively engage extinction-ensemble synapses rather than globally pruning active circuits. Local recruitment, engulfment, and protection cues determine this choice, allowing microglia to preserve fear memory by limiting extinction. HighlightsO_LIExtinction retrieval selectively recruits microglial processes to extinction ensembles. C_LIO_LIMicroglia preferentially engulf excitatory postsynaptic material from extinction ensembles. C_LIO_LILocal "find-me," "eat-me," and "dont-eat-me" cues determine synapse selection. C_LIO_LIDisrupting microglial pruning facilitates extinction without impairing fear learning. C_LI

neuroscience↗

bioRxiv2026

In vivo elastography of the human retina using light-evoked intrinsic actuation

The biomechanical properties of the retina govern its function, structural integrity, and susceptibility to disease, yet remain difficult to measure in vivo due to the lack of safe, spatially localized mechanical actuation. Here, we introduce a framework for probing retinal biomechanics in the living human eye by leveraging intrinsic optical actuation driven by phototransduction. Using phase-resolved optical coherence tomography with a local phase-referencing approach, we resolved signed, nanometer-scale displacements of the major outer retinal interfaces evoked by light. The resulting deformation field, originating in the photoreceptor outer segment, was distributed across retinal compartments in an eccentricity-dependent manner, with efficient axial transfer in the fovea and attenuated propagation in the parafovea. A hybrid analytical and finite-element framework was developed that retrieved the biomechanical properties of the retinal compartments based on their coordinated deformation and the anatomical variation in retinal structure versus eccentricity. In retinitis pigmentosa, the paradigm enabled the detection of light-evoked deformation in the transition zone despite the loss of native lamination, enabling a functional readout of the vulnerable photoreceptors at the leading edge of degeneration. Together, these results establish intrinsic optical stimulation as a basis for in vivo retinal elastography and enable the non-invasive, quantitative imaging of retinal biomechanics and function in the living human retina.

biophysics↗

bioRxiv2026

Optoretinography in R9AP-bradyopsia reveals the essential role of G-protein signaling in the human cone elongation response

Human cone and rod outer segments exhibit a rapid shrinkage followed by a slower elongation in response to light, forming the basis of the optoretinogram. The molecular basis of this optical assay of photoreceptor function remains incompletely understood. In mouse rods, the elongation response requires transducin, the G-protein -subunit activated in the initial amplifying step of the phototransduction cascade. Here, we measured human cone outer segment responses in subjects with bradyopsia arising from a triple-deletion mutation in R9AP, the anchor and transport protein for the GTPase-activating protein (GAP) RGS9. Immunoprecipitation showed that the mutant R9AP has greatly reduced affinity for RGS9, predictably reducing the level of RGS9 in the outer segment. The bradyopsia subjects elongation responses had normal activation kinetics, amplitude and photosensitivity, but markedly slowed recovery. These results indicate that normal levels of RGS9 are required for deactivation of the cone elongation response, situating the underlying molecular mechanism within the G-protein cascade at or prior to formation of the GAP complex. The recovery of the cone elongation response, measured in a paired-flash paradigm for stimuli isomerizing up to 75% of opsin, revealed that elongation requires a substrate that is depleted and recovers in a bleach-dependent manner. In particular, recovery from the highest bleaching exposure (75%) tracked the time course of cone opsin regeneration, implying that unregenerated cone opsin produces "dark light", known in rods to arise from constitutive activation of G-protein. Taken together, the results from controls and bradyopsia identify the inactive (GDP-bound) transducin complex as the essential substrate for the human cone elongation response. Suppression of the elongation response in the paired-flash paradigm further revealed the complete profile of the initial fast cone outer segment shrinkage response and reinforced its origin in opsin structural changes. Overall, these results identify key mechanistic elements of the optoretinogram and enable its use as a molecularly interpretable, non- invasive assay of cone function in disease and therapeutic response.

neuroscience↗

bioRxiv2026

EA-PheWAS: Integrating Phenotype Embeddings with PheWAS for Enhanced Gene-Phenotype Discovery

Phenome-wide association studies (PheWAS) enable systematic exploration of relationships between genetic variants and clinical phenotypes derived from electronic health records (EHRs). Conventional regression-based PheWAS treats phenotypes separately and relies on binary phenotype representations, which limits statistical power for rare variants and rare phenotypes and reduces the ability to detect associations with phenotypes that are distributed across clinical codes. To address this limitation, we first developed EmbedPheScan, a phenotype embedding-based prioritization framework that summarizes the phenotypic profiles of rare loss-of-function variant carriers in a continuous embedding space. We then proposed EA-PheWAS by combining these embedding-derived signals with conventional regression-based PheWAS results using the aggregated Cauchy association test. Using the UK Biobank whole-exome sequencing and EHR data, we show that the proposed methods maintain appropriate false-positive control. We then performed genome-wide phenome scans across all genes and across biologically defined gene classes to evaluate EA-PheWAS relative to conventional PheWAS and EmbedPheScan, consistently finding that EA-PheWAS outperformed the other two methods. We illustrate the utility of EA-PheWAS focusing on four genes representing distinct scenarios, including strong-effect disease genes (PKD1, PKD2), genes with large numbers of rare LoF carriers (NF1), and genes with extremely sparse carrier counts (FBN1).

genetics↗

bioRxiv2026

Acute exposure to cell-free mitochondrial DNA induces pregnancy-specific aortic endothelial dysfunction and organ-selective inflammation in rats

Pregnancy complications such as preeclampsia are associated with circulating cell-free mitochondrial DNA (mtDNA), a damage-associated molecular pattern capable of activating Toll-like receptor 9 (TLR9). We hypothesized that acute mtDNA exposure induces maternal inflammation and endothelial dysfunction during pregnancy via TLR9 activation. Non-pregnant and pregnant rats (gestational days 14-15) were treated intravenously with saline or purified mtDNA and euthanized 4 h after treatment. mtDNA increased cytokine mRNA expression in lung and liver of non-pregnant and pregnant rats, with magnitude varying by pregnancy status and organ. Aortas from pregnant, but not non-pregnant, rats exhibited reduced acetylcholine (ACh)-induced relaxation following mtDNA treatment (Emax, saline: 90.1 {+/-} 3.9 % vs. mtDNA: 62.1 {+/-} 20.7 % KClmax, p<0.05), while uterine artery function was preserved, indicating vascular bed-specific effects. Ex vivo incubation of aortic rings with mtDNA {+/-} white blood cells did not replicate in vivo findings, implicating systemic rather than direct vascular mechanisms. Nuclear DNA did not affect ACh-induced relaxation (p>0.05), confirming that the vascular effects were mtDNA-specific. Pharmacological antagonism of TLR9 with ODN2088 partially attenuated mtDNA-induced maternal endothelial dysfunction. Although overt vascular ROS increases were not detected, aortas from pregnant rats had reduced sod-1 expression (p<0.05) and increased eNOS protein abundance (p<0.05). Acute mtDNA exposure during pregnancy induces maternal organ inflammation and impairs endothelium-dependent vasodilation, with partial TLR9 involvement. In conclusion, aortic transcriptional changes in antioxidant pathways and increased eNOS abundance were also observed, though their functional significance remains to be determined. New & NoteworthyTo our knowledge, this is the first study to demonstrate that acute exposure to circulating mtDNA induces pregnancy-specific maternal endothelial dysfunction and organ-selective inflammatory responses. Our findings reveal pregnancy- and vascular-bed specific responses of the maternal vasculature to mitochondrial danger signals, with partial TLR9 involvement. Aortic transcriptional changes in antioxidant pathways and increased nitric oxide synthase abundance were identified as molecular correlates of this dysfunction.

physiology↗

bioRxiv2026

Three-dimensional Virtual Adult Cardiomyocyte Transcriptomics

Obtaining transcriptomes of adult cardiomyocytes at single-cell resolution remains challenging due to their large size, elongated morphology, and frequent multinucleation. Although spatial transcriptomics preserves tissue architecture and captures gene expression in situ, current analytical frameworks largely rely on nuclear-based segmentation and are therefore poorly suited to adult cardiomyocytes. Furthermore, individual tissue sections capture only a fraction of a cardiomyocyte, preventing reconstruction of complete cell-level transcriptomes. Here we present three-dimensional virtual cardiomyocyte (3D-VirtualCM), a membrane-guided framework that integrates cell-contour similarity and optimal transport to reconstruct volumetric cardiomyocyte transcriptomes from consecutive spatial transcriptomic sections. Applying 3D-VirtualCM to infarcted adult mouse hearts, we generated a panoramic transcriptomic atlas spanning 100 m thickness at single-cell resolution. 3D-VirtualCM identified spatially and transcriptionally distinct cardiomyocyte populations within the infarct border zone, enabled high-throughput quantification of cardiomyocytes re-entering the cell cycle together with their associated molecular signatures, and revealed heterogeneous RNA distribution along the longitudinal axis of individual cardiomyocytes. By integrating three-dimensional cellular morphology with in situ transcriptomic data, 3D-VirtualCM provides a scalable approach for resolving cardiomyocyte states and spatial organization in physiological and pathological cardiac remodeling.

bioinformatics↗

bioRxiv2026

DrugPlayGround: Benchmarking Large Language Models and Embeddings for Drug Discovery

Large language models (LLMs) are in the ascendancy for research in drug discovery, offering unprecedented opportunities to reshape drug research by accelerating hypothesis generation, optimizing candidate prioritization, and enabling more scalable and cost-effective drug discovery pipelines. However there is currently a lack of objective assessments of LLM performance to ascertain their advantages and limitations over traditional drug discovery platforms. To tackle this emergent problem, we have developed DrugPlayGround, a framework to evaluate and benchmark LLM performance for generating meaningful text-based descriptions of physiochemical drug characteristics, drug synergism, drug-protein interactions, and the physiological response to perturbations introduced by drug molecules. Moreover, DrugPlayGround is designed to work with domain experts to provide detailed explanations for justifying the predictions of LLMs, thereby testing LLMs for chemical and biological reasoning capabilities to push their greater use at the frontier of drug discovery at all of its stages.

bioinformatics↗

bioRxiv2026

A rugged binding landscape unifies static and dynamic paradigms in protein-protein interactions

Predicting protein-protein interaction (PPI) affinities from structural data remains a challenge. Although binding funnel theory describes the formation of native complexes, the topography of the funnel bottom and its influence on affinity are often overlooked. Using two well-controlled nanobody-antigen datasets as model systems, we demonstrate that PPI can exist in both static and dynamic binding paradigms. These two nanobody series adopt nearly identical binding poses toward their respective antigens yet exhibit diverse affinities, each representing a distinct paradigm: a static paradigm in which affinity can be ranked by Rosetta scoring of co-crystal structures alone, and a dynamic paradigm in which affinity can only be ranked using molecular dynamics-sampled ensembles. The two paradigms differ in interfacial dynamics. In the dynamic series, the relative motion between binding partners ({Delta}RMSF) is temperature-sensitive, with optimal affinity correlation at 298 K. In the static series, {Delta}RMSF is minimal and insensitive to temperature. Local frustration analysis establishes a mechanistic bridge between interfacial dynamics and landscape topology. Dynamic interfaces exhibit increased frustration upon thermal sampling, facilitating sampling of functionally relevant microstates, whereas static interfaces show minimal frustration changes across temperatures. The temperature-dependent frustration difference mirrors {Delta}RMSF sensitivity, confirming local frustration as a determinant of interfacial dynamics. Furthermore, static interfaces are characterized by a higher density of canonical hotspot residues, while dynamic interfaces utilize interfacial ruggedness to modulate affinity. Together, these results demonstrate that conserved binding modes can encode different energy landscapes and provide a practical framework for determining when ensemble-based sampling is required for accurate affinity prediction.

biophysics↗

bioRxiv2026

A feed-forward loop between niche adenosine and Gzmk⁺ CD8 T cells propagates systemic inflammaging

The causal link between the aging microenvironment and T cell aging remains elusive. Here, we demonstrate that adenosine within aging tissues actively reprograms CD8+ T cells into a pro-aging Granzyme K+ (Gzmk+) population. Mechanistically, senescent cells create an adenosine-rich niche via p16-dependent CD39 upregulation, triggering A2aR signaling to induce Gzmk+ T cell differentiation. Once released, Gzmk promotes systemic inflammaging through PAR1 and complement activation. Crucially, targeting this axis--either via genetic Gzmk ablation or pharmacological A2aR blockade--reverses multi-organ aging phenotypes and significantly extends healthy lifespan in mice. Human analysis reveals age-dependent Gzmk+ T cell accumulation in multi organs, while coffee intake (an A2aR antagonist) inversely correlates with plasma Gzmk levels. Our findings uncover how metabolic niche changes drive T cell aging and establish the adenosine-Gzmk axis as a pivotal therapeutic target for combating age-related diseases.

immunology↗

bioRxiv2026

ProteinSage: From implicit learning to explicit structural constraints for efficient protein language modeling

While protein language models typically rely on sequence-only pretraining objectives, this approach often fails to capture structural regularities and demands large datasets. To address this, we introduce ProteinSage, a pretraining framework that learns protein representations under explicit structural constraints. ProteinSage incorporates structural signals via structure-guided masking and a causal objective designed to model longrange dependencies. This structure-constrained pretraining equips ProteinSage with transferable representations using less data and computation, yet achieves competitive or superior performance across diverse structure-aware and general protein modeling benchmarks. To determine whether these gains stem from genuine structural generalization rather than task-specific fitting, we applied ProteinSage to a structure-driven protein discovery task, focusing on proteins with multi-pass transmembrane helical architectures such as distantly related microbial rhodopsins. The model successfully identified six previously unannotated microbial rhodopsin homologs. Together, our work establishes structure-constrained pretraining as an effective pathway toward data-efficient and structurally faithful protein representation learning.

bioinformatics↗