Biology subjects

Zhu, Y.

Publications and source records attributed to Zhu, Y..

At least 55 records · Page 3Linked to original sources

bioRxiv2026

Protective pan-betacoronavirus neutralizing antibodies by vaccination

The continued emergence of betacoronaviruses underscores the urgent need for vaccines that provide broadly protective immunity. Here, we present an epitope-focused vaccine strategy targeting the conserved S2 stem-helix region of the spike fusion machinery, a broadly neutralizing antibody-(bnAb) epitope shared across betacoronaviruses yet partially occluded on the native spike. Immunization of non-human primates with engineered S2 stem-helix nanoparticle immunogens, alone or followed by a SARS-CoV-2 BA.1 spike mRNA boost, elicited broadly cross-reactive antibody responses against sarbecoviruses, merbecoviruses, and embecoviruses and neutralized SARS-CoV-2, multiple variants, other sarbecoviruses, and MERS-CoV. Vaccine-elicited monoclonal antibodies displayed broad in-vitro neutralizing activity and protected against both SARS-CoV-2 and MERS-CoV in-vivo. Structural analyses revealed conserved features between rhesus and human stem-helix bnAbs, supporting the translational potential. Overall, our findings provide proof-of-concept that epitope-focused nanoparticle immunogens can target partially occluded, immunoquiescent bnAb epitopes, laying the groundwork for pan-betacoronavirus vaccines that provide broad protection and strengthen pandemic preparedness. ONE SENTENCE SUMMARYEpitope-focused S2 stem-helix nanoparticle immunogens elicit protective broadly neutralizing antibodies (bnAbs) against diverse betacoronaviruses in non-human primates, establishing a framework for development of pan-betacoronavirus vaccines.

immunology↗

bioRxiv2026

Nicotinamide N-methyltransferase couples inflammation and epigenetic remodelling to hepatic fibrosis

Background & aimsChronic inflammation is a key driver of progression from benign steatosis to metabolic dysfunction-associated steatohepatitis (MASH) and liver fibrosis. The molecular mechanisms coupling inflammatory cytokine signaling to altered hepatocyte remodeling remain incompletely understood. Here, we report that nicotinamide N-methyltransferase (NNMT) is induced by specific inflammatory cytokines in hepatocytes and acts as a key hub to connect inflammation with fibrotic liver remodeling. Approach & resultsTranscriptomic profiling of primary human hepatocytes and human hepatoma cell lines identified NNMT as a selective downstream target of IL-1{beta} and IL-6, but not of TNF. Genetic silencing of NNMT markedly attenuated cytokine-induced inflammatory and fibrotic gene expression programs and partially reversed IL-6-mediated sensitization of IL-1{beta} responses. Integration of RNA-seq with ChIP-seq, CUT&Tag, and ATAC-seq revealed that inflammatory cytokines remodel the hepatocyte epigenome through coordinated changes in histone modifications. NNMT overexpression promoted selective remodeling of H3K4me2 chromatin landscapes, accompanied by activation of fibrosis-associated transcriptional programs. Nicotinamide supplementation partially suppressed inflammatory gene expression, supporting a role of NNMT-dependent NAD+ metabolism in this process. Importantly, cytokine-induced NNMT expression and its associated transcriptional program were conserved in primary human hepatocytes and 3D liver microtissues. ConclusionsNNMT functions as an inflammatory cytokine-inducible metabolic- epigenetic integrator that links IL-1{beta} and IL-6 signaling to chromatin remodeling and inflammatory transcriptional reprogramming in hepatocytes. These findings identify NNMT as a key regulator of inflammatory and profibrotic responses during MASLD progression and highlight NNMT as a potential therapeutic target for limiting liver inflammation and fibrosis. Impact and ImplicationsThis study identifies NNMT as a previously unrecognized metabolic-epigenetic integrator that selectively links IL-1{beta} and IL-6 signaling to chromatin remodeling and inflammatory transcriptional reprogramming in hepatocytes. Our findings establish a non-canonical mechanistic framework by which inflammatory cytokines drive profibrotic gene expression during MASLD progression through. By uncovering NNMT as a critical mediator of inflammatory-epigenetic crosstalk, this work broadens our understanding of hepatocyte-intrinsic mechanisms underlying liver fibrosis and highlights NNMT as a promising therapeutic target for preventing the transition from steatosis to progressive MASH and fibrosis. HighlightsO_LINNMT is selectively induced by IL-1{beta} and IL-6 in hepatocytes. C_LIO_LINNMT controls fibrogenic gene expression modules. C_LIO_LIIL-6 amplifies IL-1{beta} responses through an NNMT-dependent mechanism. C_LIO_LINNMT links inflammatory signaling to epigenetic remodeling in MASLD. C_LI

molecular biology↗

bioRxiv2026

T50 O-GlcNAc PKM2 promotes aerobic glycolysis and PDAC progression via ARNT-CDC27-AKT pathway

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal cancers and more evidence suggests that glucose metabolism plays a significant role in the development and progression with glycosylation at multiple sites potentially being a key characteristic. However, the underlying mechanisms remain insufficiently studied. Here, we first confirmed the presence of O-GlcNAc glycosylation modifications at Thr50 on PKM, which is highly expressed in PDAC and strongly correlated with poor prognosis. Further, we found that PKM expression was significantly positively correlated with CDC27 levels, and mutation of the Thr50 O-GlcNAc site in PKM abolished the upregulation of CDC27. We confirmed that O-GlcNAc-modified PKM enhances nuclear translocation of ARNT, which binds to the CDC27 promoter to upregulate its expression. Finally, we demonstrated that reduced expression of CDC27, as a key component of the APC/C complex, leads to downregulation of ubiquitination at the K11 site of PPP2CA, resulting in upregulation of PPP2CA expression, in turn, reduces AKT phosphorylation, ultimately driving PDAC regression by inhibiting aerobic glycolysis. Thus, we delineate a novel O-GlcNAcylation-dependent pathway where PKM drives PDAC progression through ARNT-mediated CDC27 transcriptional activation and AKT-mediated glycolysis.

cancer biology↗

bioRxiv2026

Gene therapy targeting of AKAP6β-CaMKII signalosomes improves myocardial inflammation and heart failure in a swine model of cardiometabolic syndrome

BackgroundCardiometabolic heart failure with preserved ejection fraction (HFpEF) is associated with systemic and cardiac inflammation and diastolic dysfunction. A-kinase anchoring protein 6{beta} (AKAP6{beta}) is a scaffold protein located at the cardiomyocyte outer nuclear membrane that promotes pathological cardiac remodeling via the recruitment of multiple regulatory proteins including protein kinases. In mice, adeno-associated virus (AAV) mediated expression of a peptide based upon a kinase binding domain (KBD) within AKAP6{beta} inhibited the development of heart failure due to chronic pressure overload. Whether KBD expression can also inhibit the development of cardiometabolic heart failure is unknown, and if so, the mechanism of KBD action in HFpEF has yet to be explored. MethodsThe efficacy of a cardiotropic self-complementary AAV gene therapy that expresses the AKAP6{beta} KBD peptide (AAV9sc.KBD) was tested in a female Ossabaw swine model of cardiometabolic syndrome and HFpEF. Single nucleus and bulk RNA sequencing of swine heart tissue and immunoprecipitation-mass spectrometry, live cell imaging, and biochemical assays using primary rat cardiomyocytes were employed to study KBD mechanism of action. ResultsAAV9sc.KBD inhibited the development of diastolic dysfunction and heart failure in the Ossabaw model, without negatively impacting systolic function. The improvement in cardiac phenotype was associated with decreased T-cell myocardial infiltrates and partial reversal of pathological gene expression. An unbiased interactome study revealed that the KBD peptide binds Ca2+/calmodulin-dependent protein kinase II (CaMKII), identifying CaMKII as a new AKAP6{beta} binding partner. Perinuclear CaMKII activity detected by live cell imaging required AKAP6{beta} expression and was inhibited by KBD expression. In addition, the CaMKII substrate Inhibitor of NF-{kappa}B Kinase {beta} (IKK{beta}) bound AKAP6{beta}. IKK phosphorylation in the Ossabaw model and in myocytes was inhibited by KBD expression, and NF-{kappa}B nuclear translocation in myocytes was dependent upon AKAP6{beta}-CaMKII protein complex formation. AAV9sc.KBD treatment inhibited cardiomyocyte NF-{kappa}B-dependent gene expression in the Ossabaw model. ConclusionsRegulated by perinuclear AKAP6{beta}-CaMKII signalosomes, NF-{kappa}B pro-inflammatory gene expression in cardiomyocytes participates in a positive feedback loop with cardiac inflammation promoting HFpEF. Proof-of-concept is provided in a large animal model that gene therapy-based cardiomyocyte expression of the KBD peptide will prevent cardiac dysfunction in cardiometabolic syndrome. Clinical PerspectiveO_ST_ABSWhat is newC_ST_ABSO_LIThe cardiomyocyte-selective gene therapy AAV9sc.KBD, which targets signalosomes organized by the scaffold protein AKAP6{beta}, is shown to inhibit myocardial T-cell infiltration and improve cardiac structure and function in a large animal model of cardiometabolic HFpEF. C_LIO_LIThe AKAP6{beta} KBD peptide is shown to bind and inhibit the function of CaMKII. C_LIO_LICaMKII and IKK{beta} are shown to participate in perinuclear AKAP6{beta} signalosomes, where they regulate activation of the NF-{kappa}B pro-inflammatory gene regulatory pathway. C_LI Clinical implicationsO_LIProof-of-concept for a novel strategy for the treatment of HFpEF is provided, intracellular expression by a cardiomyocyte-selective gene therapy vector of an inhibitory peptide, which will inhibit compartmentalized intracellular signal transduction. C_LIO_LIIn conjunction with previous studies in small rodents, the new data obtained in Ossabaw swine support clinical translation of the AAV9sc.KBD gene therapy. C_LI

physiology↗

bioRxiv2026

Flow alters fibrin molecular and network structure and decreases binding of fibrinolytic enzymes

Thrombolysis with tissue Plasminogen Activator (tPA), approved for treating acute ischemic stroke (AIS) within 3-4.5 h of symptom onset, converts Plasminogen (Plg) to Plasmin to degrade fibrin, while fibrin also enhances Plg activation by binding to both tPA and Plg. Therefore, in this study, we investigated whether arterial-like flow, characteristic of AIS, alters fibrin structure and susceptibility to fibrinolysis. Using a newly developed platform for quantitative imaging and spectroscopy, we found that flow generates denser fibrin networks with reduced molecular transport despite reduced protofibril packing within fibrin fibers. Raman spectroscopy revealed an -helix-to-{beta}-sheet transition, accompanied by reduced Plg and tPA binding, although the reduction in tPA binding emerged only after prolonged flow exposure. Consistently, multi-scale molecular dynamics simulations showed that the Plg binding site destabilized at lower forces than the primary tPA binding site. Together, these multi-scale findings help explain the limited efficacy and narrow therapeutic window of thrombolysis.

biophysics↗

bioRxiv2026

Transplanting enzyme active site geometry into antibody CDRs for catalytic antibody design

Antibodies provide programmable molecular recognition, whereas enzymes enable repeated chemical transformation. Catalytic antibodies seek to combine these properties within a single protein scaffold. However, conventional approaches based on transition state analogue immunisation, library screening or local mutagenesis provide limited control over the atomic arrangement of catalytic residues. They also frequently produce antibodies that bind substrates without supporting efficient chemical turnover. Recent advances in generative protein design have enabled the construction of antibody complementarity determining regions and the scaffolding of functional motifs under structural constraints. A systematic strategy for transferring experimentally supported enzyme active site geometry into antibody variable domains is still lacking. Here, we present a computational framework that treats antibody and enzyme structures as distinct but complementary inputs. Developable Fv or VHH structures provide the immunoglobulin scaffold. Enzyme complexes containing substrates, products or transition state analogues provide catalytic residues, ligand conformations, metals, cofactors and key water networks. The selected catalytic atoms are mapped into antibody complementarity determining regions, while the surrounding loops are reconstructed using antibody compatible representations and constrained all atom diffusion. Sequence design and structural back prediction are followed by filters for antibody folding, catalytic geometry, ligand positioning, conformational stability and developability. The framework avoids direct fusion of intact enzymes and antibodies. Instead, it transfers only the local geometry required for catalysis. This separation of scaffold selection from catalytic motif selection creates a testable route for determining whether natural enzyme chemistry can be embedded within antibody formats. It also provides a practical basis for evaluating substrate binding, chemical conversion, product release and catalytic turnover as separate design objectives. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=91 SRC="FIGDIR/small/740676v1_ufig1.gif" ALT="Figure 1"> View larger version (45K): org.highwire.dtl.DTLVardef@2cc629org.highwire.dtl.DTLVardef@185d7e4org.highwire.dtl.DTLVardef@20f810org.highwire.dtl.DTLVardef@7e045b_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2026

Spaceland: Histology-Guided Reconstruction of High-Resolution Whole-Organ 3D Molecular Atlases from Sparse Spatial Transcriptomics

Reconstructing whole organs in three-dimensional molecular detail is a key step toward building virtual organs for modeling tissue organization, disease progression and drug perturbation responses. However, high-resolution whole-organ spatial transcriptomic profiling remains impractical, forcing a trade-off between reconstruction fidelity and sampling density. Here, we introduce Spaceland, a morphology-guided framework that reconstructs continuous, high-resolution 3D molecular landscapes from sparsely sampled spatial transcriptomic sections and serial H&E histology. Spaceland formulates this task as learning continuous gene-expression fields within a morphology-informed histological space. It constructs a dense 3D morphological scaffold by optical-flow interpolation of foundation model-derived H&E representations and decodes sparse spot-level transcriptomic measurements onto an 8 m histology-aligned grid. Across mouse olfactory bulb, mouse hemibrain and spatiotemporal planarian regeneration, Spaceland generalized across platforms, tissue scales and biological contexts. In mouse benchmarks, Spaceland bridged 400 m molecular gaps, resolved sub-spot organization, outperformed ST-based interpolation and 2D H&E-based prediction methods, and remained robust with 160-320 m H&E intervals. In planarian regeneration, it enabled time-resolved whole-organism analysis from only four Visium sections per stage, revealing dynamic neoblast-neural spatial remodeling. Together, Spaceland shifts 3D molecular atlas construction from exhaustive experimental sampling toward data-driven virtual tissue and organ modeling, providing a scalable route to whole-organ molecular reconstruction.

bioinformatics↗

bioRxiv2026

Microbial Metabolites Potentiate MAIT Cell Anti-Tumor Immunity Against Solid Tumors

Mucosal-associated invariant T (MAIT) cells sense riboflavin metabolites through the monomorphic antigen-presenting molecule MR1, providing a unique opportunity to therapeutically mobilize a broadly shared T cell population without genetic engineering. Here, we show that the highly potent microbial metabolites, 5-OP-RU and 5-OE-RU, can be exploited as pharmacologic precision immune activators to drive human MAIT cell responses against solid tumors. Ligand stimulation in human co-culture systems elicited robust MAIT cell cytotoxicity, inflammatory cytokine secretion, and transcriptional states transformation. In vivo administration of riboflavin ligands 5-OP-RU significantly suppressed tumor growth in xenograft liver cancer models. Metabolite-driven MAIT activation also reprogrammed the local immune landscape, enhancing effector function and overcoming features of the immunosuppressive niche by markedly eliminating tumor-associated macrophages (TAMs) and myeloid-derived suppressor cells (MDSCs) within the tumor microenvironment (TME). These findings reveal that microbial riboflavin metabolites can power and redirect MAIT cells to solid tumors, establishing MR1-metabolite signaling as a tractable therapeutic axis for liver cancer and other solid malignancies.

cancer biology↗

bioRxiv2026

Large-scale functional coupling and computational modelling reveal frontotemporal hotspots in distributed network connectivity during working memory recognition, encoding, and retrieval

Working memory depends on coordinated activity across distributed brain networks, but how these interactions support recognition, encoding, and retrieval remains unclear. Here, we analyzed intracranial local field potentials from 1,652 sites in 36 epilepsy patients performing a naturalistic card-matching task. Phase-amplitude and phase-phase coupling revealed widespread condition-dependent interactions, with frontal and temporal regions emerging as recurrent network hubs. Event-related analyses showed reproducible within-trial coupling dynamics across participants. Directed phase transfer entropy further identified frequency-specific information flow, including frontal regions acting as a theta-band source and an alpha-band sink. Finally, a time-delay-embedded hidden Markov model identified task-locked latent states with condition-dependent occupancy trajectories and distinct coherence profiles. Together, these results suggest that working memory is implemented through dynamic, directionally organized large-scale networks, in which frontotemporal hubs coordinate flexible transitions among distributed neural states rather than operating in isolation.

neuroscience↗

bioRxiv2026

A dual-layer computational framework for prioritising therapeutic candidates targeting extracellular vesicle-mediated immune escape in pancreatic ductal adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is an aggressive malignancy characterised by a highly immunosuppressive tumour microenvironment and limited therapeutic responses. Tumour-derived extracellular vesicles (EVs) contribute to PDAC progression by transferring immunomodulatory molecules and tumour-associated signals, suggesting EV-associated processes as potential intervention opportunities. However, the heterogeneity of EV biology and the complexity of tumour-immune interactions make single-target intervention strategies challenging. Here, we developed a computation-driven dual-layer candidate-prioritisation framework to identify potential modulators associated with PDAC EV-mediated immune escape through complementary production-side and action-side strategies. For the production-side layer, we focused on upstream processes related to EV biogenesis, cargo regulation, inflammatory signalling, and tumour-associated pathways. An 88-gene PDAC EV-associated target framework was integrated with cell-type-resolved prognosis annotations from ctPANDA and predicted targets of 18 natural products derived from Scutellaria baicalensis, Epimedium spp., and Cornus officinalis to prioritise natural-product candidates with disease relevance and potential chemical tractability. In parallel, key targets with experimentally resolved ligand-binding structures were subjected to pocket-guided de novo small-molecule design based on co-crystal ligand-defined binding sites, followed by structural, docking-based, and physicochemical screening of generated compounds. For the action-side layer, VHH and scFv binders were computationally designed and screened against extracellular regions of MET and CD81 to prioritise candidates potentially suitable for EV recognition and capture. This study provides a computational strategy for narrowing candidate spaces across both EV-associated production pathways and released vesicle recognition. The resulting small molecules, antibody-like binder models, and screening workflows provide a resource for future experimental validation of strategies targeting PDAC EV-associated immune regulation. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=118 SRC="FIGDIR/small/740294v1_ufig1.gif" ALT="Figure 1"> View larger version (48K): org.highwire.dtl.DTLVardef@11829b0org.highwire.dtl.DTLVardef@158efb5org.highwire.dtl.DTLVardef@1e17cc5org.highwire.dtl.DTLVardef@c69696_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2026

Conformational dynamics of exopolysaccharides underlie biofilm matrix mechanics in Vibrio cholerae

Polysaccharides remain the least understood biomacromolecules, particularly in terms of the relationship between their chemical structure and physical properties. On the other hand, polysaccharides often serve as the main structural components in biofilms: surface-attached aggregates of bacterial cells encased within a mechanically resilient extracellular matrix. The large chemical space explored by bacteria within biofilms provides excellent opportunities to establish the structure-function relationship for polysaccharides. In this paper, we systematically characterize various polymer properties of Vibrio polysaccharide (VPS), the major exopolysaccharide in biofilms formed by Vibrio cholerae, the causative agent of pandemic cholera. Using a combination of shear rheology, dynamic and static light scattering, and small-angle X-ray scattering, we measure the viscosity, molecular weight, persistence length, radius of gyration, and hydrodynamic radius of this chemically unique biopolymer. Combining all-atom and coarse-grained simulations, we show how the conformational flexibility of a single glycosidic linkage within each VPS monomer can lead to dramatic compaction of the entire polymer chain and nonclassical entanglement behavior. Our comprehensive quantification represents a rare endeavor for bacterial biofilms, whose matrix composition and physical properties remain largely nebulous; it also represents a significant step towards a detailed understanding of the molecular origins of biofilm mechanics.

biophysics↗

bioRxiv2026

Pervasive historical introgression across oak phylogeny shapes gene expression

Oaks were once described as the "worst-case scenario" for the biological species concept because of the presumed pervasive hybridization and introgression among species. However, we still lack a full picture of genome-wide introgression across the eight sections of the Quercus phylogeny. Furthermore, we still do not have a complete characterization of what is introgressed and how it functionally influences the recipient species. Firstly, we used 168 whole genomes from 67 Quercus species belonging to the eight sections of the Quercus phylogeny and a super pan-genome to generate a phylogeny and test for ILS and introgression. Introgression was widespread: ancient introgression occurred between ancestors of the eight sections and more recent introgression took place within sections. Secondly, using genome-wide transcriptome, methylome, and chromatin accessibility data from a subset of 19 species sampled from four sections we demonstrated that introgressed genes are more highly and uniformly expressed than non-introgressed ones. This difference in expression level between introgressed and non-introgressed genes mostly results from cis-regulation, and is most likely due to DNA methylation decay in promoter ACRs. In summary, introgression has been a major evolutionary force during the whole evolutionary history of Quercus and had a significant impact on fundamental processes such as gene regulation and expression.

evolutionary biology↗

bioRxiv2026

A structured-illumination miniscope for optically sectioned imaging and real-time neural decoding

Single-photon miniscopes enable large-scale calcium imaging in freely behaving animals but are limited by out-of-focus background fluorescence that degrades image contrast and single-cell signal fidelity. Multiphoton approaches address this limitation but remain costly and complex. Here we introduce a lightweight (<3 g), low-cost structured illumination miniscope that achieves optical sectioning in freely behaving mice. Using HiLo imaging implemented with a simple Ronchi grating and time-multiplexed excitation, the system strongly suppresses background fluorescence while preserving the speed, field of view, and accessibility of widefield miniscopes, and supports optical-sectioned multi-plane imaging to increase neuronal yield. Using hippocampal recordings, we show enhanced region-of-interest (ROI)-based signal quality and spatial information readout, allowing simple ROI-averaged signals to approach the performance of offline algorithm-extracted signals. We further demonstrate a proof-of-principle closed-loop brain-machine interface enabled by rapid online signal extraction and real-time neural decoding. Together, these results establish structured illumination as a practical and accessible strategy for achieving high-contrast calcium imaging with miniaturized microscopes.

neuroscience↗

bioRxiv2026

Cellular-Resolution Spatial Transcriptomics Reveals Laminar VSMC Phenotypic Remodeling and a Hypoxic Medial Core in Human Thoracic Aortic Dissection

Sporadic ascending aortic dissection (AAD) carries high short-term mortality and long-term morbidity. Hypertension is the predominant risk factor, yet no established tools identify patients at imminent risk. Although medial vulnerability likely contributes to dissection in heritable aortopathies, AAD predominantly understood as a luminal breach followed by false-lumen propagation, with comparatively less emphasis on the underlying medial substrate. We sought to define the vascular smooth muscle cell (VSMC) landscape in human AAD and identify spatial remodeling programs associated with interlamellar separation. Using Xenium in situ spatial transcriptomic profiling, we generated cellular-resolution maps of the dissected human ascending aorta. We identified extensive VSMC remodeling organized into distinct laminar domains across the aortic media, with distinct VSMC states supported by gene-expression module scoring and trajectory analysis. A reproducible mid-medial core of chronically hypoxia-adapted VSMCs was identified across patients and supported by carbonic anhydrase 9 immunohistochemistry. These hypoxia-adapted VSMCs lacked inflammatory and immediate-early activation programs and were spatially distinct from stress-responsive VSMC states enriched along the false lumen. Circumferential profiling in a complete aortic ring demonstrated greater adaptive remodeling in the outer curve compared with the inner curve. In contrast, donor control aortas contained fewer modulated VSMC states, less laminar striation, and no comparable hypoxia-adapted core. These findings define a spatially organized medial remodeling landscape in AAD and identify hypoxia-associated VSMC modulation as a potential feature of medial domains associated with heightened vulnerability to interlamellar separation. Defining these remodeling domains may inform improved risk stratification and experimental models that better recapitulate human AAD pathology.

genomics↗

bioRxiv2026

A multi-agent workflow converts CAR-T patient evidence into experimentally testable hypotheses

Clinical studies of chimeric antigen receptor (CAR)-T therapy generate diverse molecular and clinical evidence that remains fragmented across publications, public repositories and patient-derived datasets, limiting systematic therapeutic discovery. Here we present BioPathfinder, an evidence-guided multi-agent workflow for closed-loop biomedical discovery. BioPathfinder constructs a provenance-aware knowledge resource linking publications with patient single-cell RNA sequencing (scRNA-seq) datasets and clinical metadata, and uses role-specialized large language model agents to generate, review and prioritize diverse, falsifiable and dataset-aware mechanistic hypotheses for computational and experimental validation. Applied to a curated corpus of CAR-T-treated patient studies and matched scRNA-seq datasets, BioPathfinder identified candidate mechanisms underlying CAR-T persistence, dysfunction and therapeutic resistance. The workflow prioritized the hypothesis that genes associated with an NK-like transition programme could be targeted to reduce CAR-T exhaustion and improve persistence. Analysis of patient scRNA-seq datasets showed enrichment of this programme in exhausted post-infusion CAR-T cells. Virtual perturbation prioritized transition-associated receptor genes, including KLRC1, KLRD1 and KLRG1, and expert review selected KLRC1, encoding NKG2A, for experimental validation. In vitro and in vivo chronic-stimulation models showed that NKG2A marked activated, exhaustion-associated CD8 CAR-T cells, whereas NKG2A blockade enhanced antitumour activity and persistence-associated functional readouts in vivo. BioPathfinder establishes a generalizable framework that transforms fragmented clinical single-cell evidence into experimentally validated therapeutic hypotheses, providing a scalable strategy for AI-guided biomedical discovery. HIGHLIGHTSO_LIBioPathfinder integrates fragmented CAR-T clinical studies, patient scRNA-seq datasets and metadata into a provenance-aware evidence resource for AI-guided hypothesis discovery. C_LIO_LIA multi-agent workflow generates, critiques and prioritizes diverse, falsifiable, dataset-aware mechanisms underlying CAR-T persistence, dysfunction and therapeutic resistance. C_LIO_LIBioPathfinder prioritizes KLRC1/NKG2A as a therapeutic target, and experimental validation demonstrates that NKG2A blockade enhances CAR-T persistence and antitumour function. C_LI

bioengineering↗

bioRxiv2026

Barcode-integrated reverse transcription for accurate, complete, and low-input RNA sequencing

In conventional total-RNA sequencing, unwanted RNA is removed by ribosomal-RNA depletion or messenger-RNA enrichment, and contaminating genomic DNA is cleared by DNase, both before reverse transcription. A sample index is added only later, during amplification, so each sample is carried through reverse transcription and cleanup on its own. These steps lose material that low-input samples cannot spare, and the per-sample handling limits throughput. Here we describe BIRT (Barcode-Integrated Reverse Transcription), a hairpin primer that carries a sample barcode and begins reverse transcription preferentially at RNA 3' ends, folding both barcoding and 3'-end capture into first-strand synthesis. Barcoding at reverse transcription enables early pooling, so samples are combined before any cleanup. It also authenticates RNA against DNA contamination: the primer barcodes RNA 909-fold more often than genomic DNA, and non-barcoded reads, which report DNA, are discarded. The 3'-end preference recovers terminal sequence that random priming loses: for a typical small nucleolar RNA, 66% of reads begin within two nucleotides of the mature 3' end, against a few percent for random hexamers. We pair BIRT with PERD (Probes for Excess RNA Depletion), a modular set of blocking probes that removes unwanted RNA within the same reaction without distorting expression (Pearson r = 0.98). We have applied BIRT to 8,079 samples across seven species, from cultured cells and primary tissue to FFPE and extracellular-vesicle RNA.

molecular biology↗

bioRxiv2026

The Role of Training Paradigms in Shaping Connectivity and Dynamics: A Comparison of Gradient-Based and Evolutionary Methods in Recurrent Neural Network Models

Recurrent neural networks (RNNs) trained with backpropagation through time (BPTT) use gradients and global error signals to solve tasks, while evolutionary algorithms (EAs) offer an alternative solution through gradient-free optimization. Both classes of methods can solve the same tasks by modifying network weights, but it remains unclear how the choice of training paradigm biases the final connectivity structure and resulting dynamics. When drawing conclusions from task-trained RNNs, especially as proxies for neurobiological computation, it is important to consider whether the resulting network structure is due to the training method itself. Here, we compare four training paradigms--BPTT, evolution strategies (ES), genetic algorithms (GA), and GA combined with Ojas Hebbian plasticity rule (GA+Oja) -- across RNNs of 32, 64, and 128 neurons. These training paradigms were applied to two contrasting tasks that are exemplary of tasks which animals perform: a discrete working memory task and a continuous sensorimotor integration task. For both tasks, we analyzed how each training algorithm modified the network through four measurements: weight change allocations across input (Win), recurrent (Wrec), and output (Wout) layers, the effective rank of the recurrent weight matrix, the dimensionality of hidden-state dynamics, and task accuracy. BPTT achieved near-perfect accuracy across all task conditions, progressively allocated more weight changes to Wout as the working memory tasks difficulty increased, and confined hidden-state activity to a lower-dimensional subspace than any evolutionary method. Evolutionary methods maintained higher recurrent effective rank, higher activity dimensionality, and no comparable difficulty-dependent reallocation toward the readout, all while maintaining comparable task accuracy as BPTT. These findings show how gradient-based and gradient-free algorithms discover distinct structural and dynamical solutions to the same computational problems, especially in tasks involving working memory, with important implications for analyzing task-trained RNNs as models of biological neural computation. Biological neural circuits, which are shaped by evolution and local plasticity rather than gradient descent, may operate in higher-dimensional regimes than gradient-trained RNN models predict.

neuroscience↗

bioRxiv2026

SparseSeg: Target-Conditioned Discovery Segmentation of Cryo-Volume Electron Microscopy Under Sparse Annotation

Cryo-volume electron microscopy (cryo-vEM) enables near-native visualization of cellular ultrastructure, but its broad use is limited by low image contrast and the high cost of dense voxel-level annotation. Existing automated segmentation methods often generalize poorly across cell types, organelles, and imaging conditions. Here, we introduce SparseSeg, a target-conditioned, sparsity-driven segmentation framework that treats organelle segmentation as a discovery process rather than a closed-set classification task. SparseSeg uses a small number of context-specific exemplars to iteratively propagate reliable supervision through the volume. It combines sparse patch-based sampling, a multi-kernel U-Net, and geometry-consistent refinement to expand accurate segmentation while suppressing context-dependent false positives. Across serial cryo-FIB-SEM and conventional vEM datasets, SparseSeg achieves robust segmentation under extreme sparse annotation, including settings with less than 1% labeled slices. This framework reduces annotation burden while preserving morphological fidelity for quantitative cryo-vEM analysis.

bioengineering↗