Biology subjects

Cai, Y.

Publications and source records attributed to Cai, Y..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

Chemical inactivation of a bacterial immune system de-domesticates a temperate phage and promotes its spread

Microbial communities shape hosts and ecosystems, yet the behaviors of these microbes are themselves controlled by mobile genetic elements. These mobile genetic elements (MGEs) compete and influence one anothers distributions. For example, many anti-phage immune systems are carried by prophages, plasmids, and integrative and conjugative elements (ICEs). It is unclear how important each of these interference mechanisms is for the spread of different mobile elements. Chemical inhibitors of these mechanisms of inter-MGE competition can clarify their mechanisms and significance in diverse environments. We discovered chemical inhibitors of SpbK--an antiphage defense encoded by an ICE in several strains of Bacillus subtilis. It is believed to prevent the dissemination of SP{beta} -like temperate phages, forcing SP{beta} to remain domesticated in its ICE-containing host. Chemical inhibition of SpbK dramatically improved the escape of infectious SP{beta} phages capable of disseminating and forming new lysogens. Furthermore, SpbK is part of a widespread class of Toll/interleukin-1 receptor (TIR)-domain containing enzymes that are prevalent both in prokaryotic antiphage immune systems and eukaryotic immune signaling. Our inhibitors reveal two distinct mechanisms of inhibiting TIR-domain enzymes. These models of inhibition may broadly apply across immune proteins, enabling inhibition of many TIR immune systems that influence phage spread as well as plant immunity and animal cellular signaling.

biochemistry↗

bioRxiv2026

Complete biosynthesis of penicillin G in Nicotiana benthamiana

Commercial penicillin production has relied on microbial fermentation for more than 80 years. Here, we engineered the plant, Nicotiana benthamiana, to produce penicillin G in its leaves by transient expression of up to seven fungal biosynthetic genes. Remarkably, all recombinant proteins localize to the analogous subcellular compartments without engineering signal peptide sequences or post-translational modification sites. Although non-ribosomal peptide synthetases occur widely in fungi and bacteria to produce a plethora of specialized metabolites, their evolutionary distribution does not extend to plants. Our results now open a new metabolic frontier for natural product synthesis, and offer possibilities to address global health concerns through an alternative biotechnology platform for fungal-derived pharmaceutical production.

synthetic biology↗

bioRxiv2026

Tree diversity intensifies soil microorganism-tree interactions

O_LIA productivity-driven higher nutrient demand of trees in diverse mixtures is frequently reported. Yet, it remains unclear how tree diversity influences microorganisms-plants interactions, in which microbes facilitate tree nutrient acquisition in exchange for carbon (C) to meet the resource demand of both. C_LIO_LIUsing a long-term tree diversity experiment in the subtropics, we assessed microbial investment in C-, nitrogen (N)-, and phosphorus (P)-acquiring enzymes in litter and mineral soil, testing the effects of tree species richness and mycorrhizal type (arbuscular (AM)- vs. ectomycorrhizal (EcM)-associated tree species). C_LIO_LIWith increasing tree species richness, microbial investment in C acquisition decreased, while investment in N and/or P acquisition increased in litter and in mineral soil. In mineral soil of AM-associated tree mixtures, ecoenzymatic stoichiometry revealed a shift from microbial investment in C toward P acquisition as tree species richness increased. C_LIO_LIOur findings suggest that tree diversity strengthens microbe-tree interactions in terms of C-for-nutrient exchange. This highlights the key role of soil microorganisms, particularly in AM symbiosis, shaping tree diversity-biogeochemical feedbacks. C_LI

ecology↗

bioRxiv2026

Lamellar Normative Modelling of the Hippocampus Across the Human Lifespan

The hippocampus is a central hub of human memory and cognition and is closely associated with brain disorders. Studies have shown that it exhibits complex structural variation across the lifespan, yet the details of hippocampal morphology changes remain poorly understood. Here, we establish norms over the hippocampal geometry that resolve lamellar morphology and map lifespan trajectories across more than 27,000 individuals from 158 scanning sites. Hippocampal geometry shows spatially non-uniform developmental and ageing patterns across the lifespan, with lamellar thickness, width and length following dissociable trajectories. Across multiple brain disorders, this representation reveals localized and heterogeneous alterations beyond conventional subfield-level summaries, and uncovers a dichotomy in disease-associated patterns, with neurodegenerative conditions and schizophrenia showing predominant atrophy, whereas some other disorders exhibit focal or regionally selective hypertrophy. Transfer to a longitudinal Alzheimers Disease Neuroimaging Initiative cohort further supports out-of-sample generalization of our approach and enables individual-level tracking and conversion risk stratification. Overall, this work establishes a population-scale geometric reference for the hippocampus, extends normative brain mapping from coarse regional phenotypes to anatomically organized subcortical structure, and enables anatomically grounded characterization of disease-related alterations and individual-level deviation mapping, providing a principled basis for understanding and stratifying brain disorders across the lifespan in health and disease.

neuroscience↗

bioRxiv2026

Dark adaptation of cone photoreceptor responses is revealed by optoretinography

Dark adaptation is the essential process that restores visual sensitivity following exposure to bright light, yet the underlying mechanisms remain incompletely understood. Here, we propose a method for assessing dark adaptation in cones using optoretinography (ORG) based on adaptive optics optical coherence tomography (AO-OCT). ORG quantifies cone functional response by monitoring nano-scale changes in the cones outer segment occurring over hundreds of milliseconds after visible stimulation. This method consists of sequential measurements of stimulus-evoked cone responses over the course of minutes of dark adaptation. Each response captures optical path length changes in single photoreceptor outer segments over milliseconds during a multi-minute recovery period following a strong photopigment bleach. We parameterized cone ORG responses and proposed an exponential model linking ORG dynamics to pigment regeneration. Parameters of the ORG response exhibited exponential decay behavior during dark adaptation, and were thus fit with exponential functions and quantified by the resulting decay parameter{tau} . Parameters capturing the amplitude of the ORG responses recovered more slowly than those capturing temporal dynamics of the responses. This difference is consistent with distinct contributions from photopigment regeneration and downstream phototransduction processes. Recovery speed varied by two-to threefold among three normal subjects, suggesting substantial inter-subject physiological diversity. Processes within the cone, including pigment regeneration, are thought to underlie the gains in photopic visual sensitivity that occur in the dark. These findings highlight ORG as an objective and sensitive assay of those cellular mechanisms. While the ORG itself has shown promise as a biomarker of the health of the photoreceptor response to light, the results of this study show that it may also be useful for probing the health of the intra- and intercellular homeostatic mechanisms that support it.

neuroscience↗

bioRxiv2026

Chemical suppression of a bacterial immune system revives repressed phages

Many antiviral immune systems have recently been discovered in bacteria. The mechanisms of several are obscure, as are their individual significance for antiphage defense. To shed light on the mechanism and significance of the two-component type I Thoeris antiphage immune system, we leveraged high-throughput phenotypic screening to identify three small molecule inhibitors. The inhibitors target the ThsA NADase component, inhibiting its 3'-cADPR-activated filamentation. The temporal control afforded by the small-molecule inhibitors allowed us to answer an outstanding question in antiviral immunity--is persistent immunity required to repress phage titers, or do immune systems become unnecessary after eradicating infectious phages? We found that Thoeris immunity must be maintained, as chemical inhibition enabled repressed phages to revive and overtake the bacterial population. Furthermore, due to the cooperative nature of antiviral immunity, we found that Thoeris must be inhibited in only 10% of the bacteria to cause phage-induced lysis of the entire population.

biochemistry↗

bioRxiv2026

Engineered Vibrio natriegens lysate can replace multiple components of cell culture media

Reducing the cost and environmental impact of cell culture media is an important goal for cultivated meat, the process of generating meat in vitro using proliferating animal cells. While prior approaches have demonstrated the use of microbial lysates to replace expensive animal-based fetal bovine serum (FBS) in media, these formulations still rely on large quantities of growth factors such as fibroblast-like growth factor 2 (FGF2). Here, we demonstrate the use of FGF2-expressing Vibrio natriegens to create whole-cell lysates that replace both FBS and FGF2 in cell culture media for cultivated meat applications. This medium, named "VN40FGF", supports rapid proliferation of immortalized bovine muscle satellite cells (iBSCs) in the absence of supplemented FGF2. Cells grown in VN40FGF maintain phenotype and differentiation capacity. We also demonstrate that engineered V. natriegens can grow in spent cell culture media, further improving sustainability and economics, and reducing potential eutrophication concerns associated with waste disposal. Our approach combines multiple strategies for reducing the total number of media inputs, demonstrating opportunities for more economical and sustainable cell culture, especially for cultivated meats.

bioengineering↗

bioRxiv2026

IEKB: a comprehensive knowledge base for inner ear genetics integrating curated associations, cochlear interactions, Bayesian candidate prioritisation, explainable dark-gene support relations, and a scientific entity network

Inner-ear genetics has expanded rapidly, yet the supporting evidence remains dispersed across a vast literature and across resources that typically emphasise loci, variants, or expression data rather than integrated biological interpretation. Here we present the Inner Ear Knowledge Base (IEKB; https://earkb.org), an open database that unifies curated associations, cochlear interaction evidence, candidate prioritisation, explainable support relations, and network exploration for inner-ear research. IEKB was built with an automated agent-assisted curation workflow that combines schema-constrained literature extraction, continuous human monitoring, and final expert review by inner-ear genetics researchers. By systematically analysing 250,696 PubMed-indexed records retrieved across 16,563 screened genes, IEKB curates 6,051 gene-phenotype-disease associations from 2,494 genes across 43 phenotype categories and 4,102 cochlear gene-gene interactions with pathway, cell-type, and experimental context. IEKB further includes a Bayesian "dark matter" module that prioritises 243,071 candidate gene-phenotype associations for 13,229 genes across all 43 phenotypes (global AUC-ROC = 0.8603; global AUC-PR = 0.1674), together with a supervised dark-relation layer that ranks phenotype-specific known-gene support for each candidate and a multi-entity scientific network containing nearly 4,000 entities, 28,616 deterministic edges, and 83,712 literature-derived relational links. The web resource supports interactive search, multi-parameter filtering, gene-detail pages, bibliometric exploration, domain-specific enrichment against IEKB phenotype and disease gene sets, network visualisation, bulk download in CSV, JSON, SQLite, and XLSX formats, and natural-language evidence-grounded question answering through a companion conversational interface (IEKB QA). To our knowledge, IEKB is the first openly accessible inner-ear resource that integrates curated associations, cochlear interactions, probabilistic candidate prioritisation, auditable known-gene support relations for novel candidates, and a multi-entity scientific network within a single database. All data are released without registration under the CC BY 4.0 license.

bioinformatics↗

bioRxiv2026

Ramping-up hippocampal ripples and their neocortical coupling support human visual short-term memory

Emerging evidence suggests that the hippocampus contributes to visual short-term memory (VSTM). However, the neural mechanisms through which hippocampal activity supports the maintenance of VSTM representations remain largely unknown. Here, using intracranial EEG recordings from human participants performing a delayed match-to-sample task for naturalistic objects, we investigated the role of hippocampal ripple activity--brief high-frequency oscillations associated with memory replay--in VSTM maintenance and hippocampal-neocortical communication. We found that hippocampal ripple rates progressively ramped up during the maintenance period and supported successful VSTM. More critically, hippocampal ripples were temporally coupled with the ripples in the lateral temporal lobe (LTL), and these coupled ripples were associated with memory reactivation in the LTL. These findings provide direct evidence that hippocampal-neocortical interaction via coupled ripples supports VSTM, extending the role of hippocampal ripples from long-term consolidation to short-term mnemonic processes and offering new insights into the dynamic coding mechanisms of working memory. Impact StatementHippocampal ripple ramping-up and neocortical coupling support a dynamic coding mechanism for visual short-term memory maintenance

neuroscience↗

bioRxiv2026

Exploring Chromosomal Position Effects for Predictable Tuning of Metabolic Pathways in Yeast

Precise control of gene expression is essential for optimizing metabolic pathways, yet current tuning strategies based on promoter strength or gene copy number remain largely empirical. Chromosomal position represents an additional regulatory axis, as identical gene expression cassettes can exhibit markedly different expression levels depending on their integration sites. Here, we systematically quantified the expression output of 98 intergenic regions (IGRs) in Saccharomyces cerevisiae using a fluorescent reporter and developed a predictive framework, Yeast IGR Prophet (YeIP), that infers expression potential directly from genomic context. By integrating multi-scale genomic features including transcriptional neighborhood, chromatin state, and chromosome topology, YeIP accurately predicted expression ranks and enabled the construction of a genome-wide atlas of expression hotspot and coldspot. Using this atlas, we rationally optimized a three-gene lycopene pathway solely through genomic integration site selection, achieving optimal transcriptional stoichiometry without modifying promoters or gene copy numbers. These results transform chromosomal integration sites from static genomic coordinates into programmable regulatory elements, establishing a predictive, data-driven framework for rational and scalable design of metabolic pathways in yeast.

synthetic biology↗

bioRxiv2026

A Long-Context Generative Foundation Model Deciphers RNA Design Principles

Programmable design of RNA sequences with defined functions remains a central challenge in biology. Despite recent advances, existing RNA generative models lack robust controllable design capabilities and are constrained by short context windows, limiting their capacity to model the complex evolutionary manifold of full-length transcripts. Here, we introduce EVA (Evolutionary Versatile Architect), a long-context generative RNA foundation model trained on over 114 million full-length RNA sequences spanning the full breadth of evolutionary diversity. By integrating a Mixture-of-Experts architecture with an 8,192-token context window, EVA learns a unified, evolution-consistent representation of RNA sequence space that captures both global transcript structure and fine-grained functional elements. EVA establishes a unified framework for a broad spectrum of RNA design tasks within a single architecture, ranging from mutational fitness prediction to context-aware sequence engineering. We demonstrate EVAs versatility through the de novo design and targeted optimization of diverse RNA classes, including tRNAs, aptamers, and CRISPR guide RNAs, as well as mRNA and circular RNA therapeutics. In systematic evaluations, EVA achieves state-of-the-art performance across seven of nine established benchmarks, improving structural modeling accuracy by up to an order of magnitude compared to existing approaches. EVA is fully open-source at the following repository: https://github.com/GENTEL-lab/EVA.

molecular biology↗

bioRxiv2026

LysinFusion: Integrating Multi-Feature Encoding and Hybrid CNN-Transformer Architecture for Phage Lysin Prediction

The escalating antimicrobial resistance crisis has intensified the urgent need for alternative antibacterial agents. Phage lysins exhibit potent bactericidal activity with low resistance potential, yet their large-scale discovery from rapidly expanding genomic resources remains limited, as existing computational methods often rely on sequence homology and few reproducible, independently validated, and practically deployable prediction frameworks are currently available. We present LysinFusion, a reproducible deep learning framework that integrates heterogeneous sequence features via a hybrid CNN-Transformer architecture for accurate lysin identification. Trained on a curated, de-redundant corpus (PHROG + inphared) and validated on 148 experimentally confirmed UniProt proteins, LysinFusion achieves an accuracy of 0.81, an AUC of 0.89, and an MCC of 0.62, outperforming the state-of-the-art DeepMineLys by up to 52.3% across metrics while reducing false positives by 64% (12 vs. 33). Ablation studies confirm that both CNN and Transformer modules are individually essential and that serial processing outperforms parallel alternatives. Model interpretability via occlusion and LIME aligns with established lysin biology, with attention concentrating on the N-terminal catalytic domain and decision boundaries reflecting characteristic charge-composition thresholds documented in Gram-negative lysin C-terminal regions. The framework, including all source code, curated datasets, and an accompanying web server, is freely available at https://github.com/sinuo560/LysinFusion.

bioinformatics↗

bioRxiv2026

MERFISH 2.0, an ultra-sensitive single-cell spatial transcriptomics imaging chemistry across diverse tissue types

Spatial transcriptomics has emerged as a transformative approach for elucidating tissue architecture, cellular heterogeneity, and disease mechanisms by preserving the spatial context of gene expression in cells. Despite these advances, many spatial transcriptomic methods underperform in archival or clinically relevant specimens, particularly formalin-fixed, paraffin-embedded (FFPE) tissues, where RNA degradation and crosslinking hinder transcript detection. To address these challenges, we developed Multiplexed Error Robust Fluorescence In Situ Hybridization 2.0 (MERFISH 2.0), an optimized spatial transcriptomic imaging chemistry to enhance profiling of fragmented and highly crosslinked RNA. Across diverse human and mouse tissues preserved as fresh-frozen, fixed-frozen, and FFPE specimens, MERFISH 2.0 substantially increased transcript detection sensitivity by up to [~]8-fold relative to MERFISH 1.0 while preserving quantitative concordance (Pearson r [≥] 0.8 across tissues). In archived fresh-frozen human brain samples, MERFISH 2.0s enhanced sensitivity improved transcript recovery, enhanced cell type resolution and spatial analyses. In low quality archival FFPE human breast cancer specimen, MERFISH 2.0 revealed additional cell populations, novel cell clusters, refined tumor-immune architecture, and increased detection of gene-gene and cell-cell interactions relative to MERFISH 1.0, underscoring the impact of improved sensitivity on downstream spatial analysis. By substantially expanding robust transcript detection to degraded and archival samples, MERFISH 2.0 enables scalable, cohort-level spatial transcriptomic analysis across clinically relevant tissue collections.

genomics↗

bioRxiv2026

TargetMITO: A rule-based model for generating highly functional synthetic mitochondrial targeting sequence in yeast

Mitochondria are essential organelles containing their own genomes, encoding a few proteins essential for energy production. Most of the mitochondrial proteins are nucleus-encoded, translated as precursors in the cytoplasm, with a large fraction of these precursors properly addressed by an N-terminal mitochondrial targeting sequence (MTS). These MTS share common features but no consensus sequence can explain their functionality nor the precursors-specific determinants of mitochondrial import. To decipher this mechanism, we created a simple computational model to generate highly functional synthetic MTS while maintaining a tight control on the design parameters. Using the budding yeast, we demonstrated the presence of precursors-specific signatures in addressing artificially nucleus-relocated OXPHOS proteins. We also show the ability of six promising candidate synthetic MTS to address a fluorescent reporter to human mitochondria cells. Our research work confirms the uniqueness of the MTS-passenger protein synergy and takes us one step closer towards improving gene therapy-based treatment of mitochondrial diseases. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=81 SRC="FIGDIR/small/707306v1_ufig1.gif" ALT="Figure 1"> View larger version (27K): org.highwire.dtl.DTLVardef@ab905forg.highwire.dtl.DTLVardef@19365eforg.highwire.dtl.DTLVardef@956b9org.highwire.dtl.DTLVardef@11cb7cf_HPS_FORMAT_FIGEXP M_FIG C_FIG

synthetic biology↗

bioRxiv2026

In vivo deuteration reveals pronounced variation in myelin lipid turnover rates and reduced myelin renewal with ageing

Myelin turnover is essential for its structural and functional integrity, yet how this particularly lipid-rich membrane is renewed and why it deteriorates with ageing remain unresolved. Combining deuterium oxide administration in mice with high resolution lipidomics, we establish that brain lipid turnover rates are highly heterogeneous, differ by brain region, and depend primarily on lipid class. Half-lives of common glycerophospholipids in purified myelin were under 2 months whereas many sphingolipids exhibited half-lives exceeding 8 months, dependent on acyl chain length and saturation. Myelin sphingolipid and cholesterol replacement rates in the corpus callosum decreased markedly between 3 and 12 months of age, while disrupting lipid trafficking through ApoE ablation preferentially impaired cholesterol turnover and incorporation into myelin. Our results establish that myelin renewal occurs through continual replacement of individual lipid constituents in a manner that depends on lipid class, hydrophobicity, and ApoE-dependent trafficking, and that this process slows significantly with ageing.

biochemistry↗

bioRxiv2026

Organoid Profiler: Automated, high-throughput and quantitative morphological characterization uncovers conserved longitudinal developmental kinetics in microfluidics-engineered organoids

Three-dimensional organoid systems have emerged as transformative tools in biomedical research, yet their translation into standardized high-throughput screening platforms is impeded by stochastic fabrication variability and a lack of scalable quantitative analytics. Here, we present a unified framework integrating microfluidic droplet engineering with Organoid Profiler, an automated deep-phenotyping pipeline designed to capture longitudinal morphological kinetics. By processing over 10,000 longitudinally tracked images across murine lung and liver and human cerebral models, we show that this platform achieves human-level segmentation precision (r = 0.99) while uncovering conserved developmental dynamics obscured in manual cultures. We identify a distinct biphasic "remodeling-to-expansion" trajectory in microfluidic organoids, characterized by an initial compaction phase underpinned by the transcriptional upregulation of focal adhesion pathways, followed by exponential expansion. We further demonstrate a "shape relaxation" phenomenon where tissues converge to spherical equilibrium regardless of initial geometry, indicating intrinsic self-organization. Multi-modal profiling confirms that microfluidic organoids exhibit superior viability, homogeneity (area variability [≤] 5% upon fabrication, [≤] 15% across development) and biological fidelity, retaining physiological levels of immune and stromal niche components often lost in conventional culture. This study establishes morphological metrics as reliable, non-invasive proxies for tissue health and molecular maturity, providing a standardized foundation for quantitative organoid biology.

bioengineering↗

bioRxiv2026

TMEM33 deletion potentiates anti-tumor CD8+ T cell immunity

Improving responses to cancer immunotherapies requires deeper insight into the cellular mechanisms governing T cell-mediated anti-tumor immunity. TMEM33 is an endoplasmic reticulum-resident transmembrane protein enriched across multiple tumor types, with reported functions in anti-viral immunity as well as calcium and lipid homeostasis, yet its role in tumor immunosurveillance remains unknown. Using murine genetic models, we demonstrate that host TMEM33 constrains anti-tumor CD8+ T cell responses. Constitutive Tmem33-/- mice exhibited delayed melanoma tumor growth and increased CD8+ T cell infiltration. Antigen-specific CD8+ compartments in tumors of Tmem33-/- mice showed TCF-1+PD-1+ progenitor-exhausted cell (Tpex) enrichment, elevated effector function and reduced exhaustion, alongside improved effector memory expansion and T-bet expression in draining lymph nodes. We highlight that TMEM33 functions intrinsically within the T cell compartment, as TMEM33 deletion (1) enhanced polyclonal activation of naive CD8+ T cells ex vivo, (2) promoted preferential Tpex accumulation among adoptively transferred naive OT-I cells in B16F10-OVA tumors and draining lymph nodes, and (3) improved the potency of ex vivo-expanded OT-I cells in controlling tumor growth during adoptive cell therapy. Finally, in a large, prospectively recruited metastatic melanoma cohort, lower TMEM33 expression in patient CD8+ T cells significantly correlated with improved survival and elevated TCF-7 (encoding TCF-1). Collectively, our findings define TMEM33 as a formerly unrecognized intrinsic determinant of tumor-directed CD8+ T cell fate that limits Tpex maintenance, and restrains cell therapy responses, suggesting that its modulation may strengthen immunotherapeutic efficacy. One sentence summaryTMEM33 intrinsically limits progenitor exhausted CD8+ T cells, scales anti-tumor responses and predicts melanoma patient survival.

immunology↗

bioRxiv2025

Understanding the LLM-based gene embeddings

Large language model (LLM)-derived gene embeddings, generated from brief NCBI gene descriptions, have shown strong performance in recent biological applications, yet the biological information they contain remains unclear. We evaluate these embeddings using a Gene Set Enrichment Analysis (GSEA)-based framework that treats each embedding dimension as a potential carrier of pathway-level information. OpenAIs embeddings recover over 93% of Hallmark and C2 pathways, with pathway signals distributed across many dimensions. Even embeddings generated from gene symbols alone recover more than 64% of pathways, indicating substantial prior biological knowledge embedded in the model. Comparing 11 small language models reveals that domain-specific models perform best with minimal input, but all models approach OpenAI-level coverage when given modest textual context. Collectively, these results show that LLM-derived embeddings encode unexpectedly extensive pathway-level information, supporting their use as lightweight, informative representations for downstream biological analysis.

bioinformatics↗