Biology subjects

Ji, Z.

Publications and source records attributed to Ji, Z..

At least 37 records · Page 2Linked to original sources

bioRxiv2025

Systematic microcircuit reconfigurations underlie early experience-induced visual cortical plasticity

Experience-dependent reorganization of neuronal connections during a postnatal critical period (CP) is essential for the functional maturation of developing primary visual cortex (V1). However, the systematic reconfigurations across microcircuit synapses underlying this plasticity were poorly defined. Paired recording from excitatory pyramidal cells and other three major interneurons (INs) within layers 2/3 and 4 microcircuits in the developing mouse V1 monocular region, we find that monocular deprivation (MD) for one-day during the CP transiently reconfigures local excitatory synapses on inhibitory INs that specifically express parvalbumin (PV) or vasoactive intestinal peptide (VIP), whereas prolonged MD primarily strengthens inhibitory transmission from somatostatin (SST)-expressing INs to PV-INs and VIP-INs in both layers. A physiologically based microcircuit model suggests these coordinated synaptic changes in specific inhibitory neurons mediate differential MD-induced plasticity in visual responses across layer 2/3 neurons in vivo. Together, our findings define a systematic microcircuit basis for cortical critical-period plasticity. HighlightsO_LIInhibitory-to-inhibitory connectivity is conserved in cortical layer 2/3 and 4 circuits. C_LIO_LIEarly experience sequentially modifies excitatory and dis-inhibitory synapses in both layers. C_LIO_LISystematic microcircuit reconfigurations underlie plasticity of cortical visual responses. C_LIO_LIModeling suggests a pivotal role of disinhibitory microcircuits in mediating the plasticity. C_LI

neuroscience↗

bioRxiv2025

Torsins Organize CLCC1 Assembly to Safeguard ER Bilayer and Lipid Homeostasis

The TMEM41B scramblase and its regulatory partner CLCC1 initiate lipid flux by equilibrating newly-synthesized phospholipids across the endoplasmic reticulum (ER) bilayer, a fundamental process required for diverse events ranging from membrane biogenesis to bulk lipid supply. Loss of CLCC1/TMEM41B causes ER bilayer imbalance, inducing giant ER-enclosed lipid droplets (geLDs) and driving rapid progression into severe MASH. Here we identify CLCC1 as the long-missing client of the lumenal Torsin ATPases, which selectively engage oligomerized CLCC1 at sites of ER bilayer imbalance. Hepatic TorsinA inactivation triggers geLD formation amid disrupted lipoprotein biogenesis and severe MASH, closely phenocopying CLCC1/TMEM41B deficiency. Mechanistically, Torsins act as foldases that drive CLCC1 oligomerization for its recruitment to imbalanced bilayers. Remarkably, ectopic CLCC1 expression reverses cellular and systemic lipid disorders arising from hepatic TorsinA deficiency. Hence, Torsin ATPases emerge as fundamental regulators that organizes CLCC1 and the downstream TMEM41B scramblase to govern lipid partitioning and membrane homeostasis.

cell biology↗

bioRxiv2025

Pseudotimecascade visualizes gene expression cascade in pseudotime analysis

Single-cell transcriptomic technologies enable the reconstruction of dynamic biological processes such as cell development and differentiation. While existing pseudotime methods allow the analysis of temporal expression patterns, they primarily focus on individual genes, overlooking the coordinated programs that drive cellular transitions. We introduce Pseudotimecascade, a tool for visualizing and comparing multi-gene expression cascades along pseudotime. In addition, it links these cascades to biological functions by identifying stage-specific pathways. Applied to hematopoietic stem cell differentiation, Pseudotimecascade highlights regulatory hierarchies and stage-specific processes, offering a deeper understanding of gene programs that govern cell fate decisions. The Pseudotimecascade package is freely available at https://github.com/changxinw/Pseudotimecascade.

bioinformatics↗

bioRxiv2025

Unfolding the Black Box of Recurrent Neural Networks for Path Integration

Path integration is essential for spatial navigation. Experimental studies have identified neural correlates for path integration, but exactly how the neural system accomplishes this computation remains unresolved. Here, we adopt recurrent neural networks (RNNs) trained to perform a path integration task to explore this issue. After training, we borrow neuroscience prior knowledge and methods to unfold the black box of the trained model, including: clarifying neuron types based on their receptive fields, dissecting information flows between neuron groups by pruning their connections, and analyzing internal dynamics of neuron groups using the attractor framework. Intriguingly, we uncover a hierarchical information processing pathway embedded in the RNN model, along which velocity information of an agent is first forwarded to band cells, band and grid cells then coordinate to carry out path integration, and finally grid cells output the agent location. Inspired by the RNN-based study, we construct a neural circuit model, in which band cells form one-dimensional (1D) continuous attractor neural networks (CANNs) and serve as upstream neurons to support downstream grid cells to carry out path integration in the 2D space. Our study challenges the conventional view of considering grid cells as the principal velocity integrator, and supports a neural circuit model with the hierarchy of band and grid cells.

neuroscience↗

bioRxiv2025

Mutations in the ubiA gene are the major mechanism of ethambutol resistance in Mycobacterium avium

Mycobacterium avium-intracellulare complex (MAC) infections are the most common nontuberculous mycobacterial (NTM) infections, and ethambutol (EMB) is one of the main therapeutic agents used to treat MAC infections. However, the EMB resistance profile and its resistance mechanisms in M. avium remain poorly understood. This study determined the minimum inhibitory concentration (MIC) on 40 M. avium clinical strains, revealing that 97.5% (39/40) of the strains were intermediate (MIC=4 g/mL) or resistant (MIC[≥]8 g/mL) to EMB. One susceptible clinical isolate strain 245 (MIC=2 g/mL) was picked for EMB-resistant mutant isolation, and a total of 121 resistant mutants were isolated and subjected to whole-genome sequencing or Sanger sequencing. Integrated analysis revealed that 94.21% (114/121) of the mutants carried mutations in the ubiA gene, which encodes DPPR (decaprenylphosphoryl-{beta}-D-5-phosphoribose) synthase - an enzyme involved in cell wall biosynthesis that has been associated with high-level EMB resistance in M. tuberculosis. Complementation with the wild-type ubiA gene restored EMB susceptibility in EMB-resistant mutants and clinical strain 322 (EMB MIC=64 g/mL) with ubiA mutation, reducing the MICs from 32 g/mL to 4 g/mL and from 64 g/mL to 8 g/mL, respectively. This study indicates that the ubiA mutation is the major mechanism of EMB resistance in M. avium, which should facilitate the development of molecular tests for rapid detection of resistance in this organism.

microbiology↗

bioRxiv2025

Hippocampal theta sweeps indicate goal direction

Successful spatial navigation requires rapid evaluation of potential future trajectories. Hippocampal "theta sweeps", sequential activation of place cells within individual theta cycles, exhibit predictive dynamics within the ideal timeframe to fulfill this role. However, whether these sequences simply reflect movement-related variables or afford more cognitive goal-directed planning remains unresolved. Using data from a navigation task on the "Honeycomb" maze that allows dissociation of head-, movement- and goal-direction correlates, we found that hippocampal theta sweeps exhibit robust goal-oriented directional biases, independent of movement- or head-direction. An existing model of theta sweeps, with an additional goal-oriented directional input, reproduces these findings and predicts goal-oriented theta phase precession, which we confirm empirically. Replay events during immobility-related sharp wave/ripples are also goal-directed, and therefore more aligned with theta sweeps than experience. Our findings indicate that hippocampal theta sweeps provide a neural substrate for online goal-directed spatial planning.

neuroscience↗

bioRxiv2025

Evaluating large language models in biomedical data science challenges through a classroom experiment

Large language models have shown remarkable capabilities in algorithm design, but their effectiveness in solving data science challenges remains poorly understood. We conducted a classroom experiment in which graduate students used large language models (LLMs) to solve biomedical data science challenges on Kaggle. While their submissions did not top the leaderboards, their prediction scores were often close to those of leading human participants. LLMs frequently recommended gradient boosting methods, which were associated with better performance. Among prompting strategies, self-refinement, where the LLM improves its own initial solution, was the most effective, a result validated using additional LLMs. These findings demonstrate that LLMs can design competitive machine learning solutions, even when used by non-experts.

bioinformatics↗

bioRxiv2025

The XPF-like domain in SHOC1 required for homologous recombination and safeguarding autosome from meiotic silencing of unsynapsed chromatin

During meiosis, a group of evolutionarily conserved ZMM proteins plays essential roles in stabilizing the recombination intermediates and promoting crossover (CO) formation. In mice, SHOC1 forms a trimeric complex with the other two ZMM proteins, SPO16 and TEX11, to bind recombination intermediates after strand invasion. Although genetic variants of SHOC1 are clinically associated with meiotic arrest and male infertility, their precise molecular mechanisms and evolutionarily conserved functions in human gametogenesis remain enigmatic. Here, we delineated species-specific divergences between human and mouse SHOC1 complex, and identified a missense variant within the XPF-like domain in SHOC1 (c.A1769G:p.Q590R) that was associated with meiotic arrest and non-obstructive azoospermia (NOA). The disorder of the XPF-like domain in SHOC1 impaired DNA double-strand breaks repair by compromising its ability to bind branched DNA structures and the recruitment of M1AP, REDIC1, and ZMM factors to recombination intermediates, ultimately abolishing CO formation. Furthermore, the variant disrupted dynamic 3D chromatin structure in pachytene spermatocytes and induced defects in homologous chromosome synapsis. More importantly, the XPF-like domain in SHOC1 was revealed to prevent autosome intrusion into the sex body compartment, thereby safeguarding critical autosomal loci from meiotic silencing of unsynapsed chromatin (MSUC). Overall, our study demonstrated that the XPF-like domain in SHOC1 is required for homologous recombination and safeguarding autosome from MSUC in meiosis.

developmental biology↗

bioRxiv2025

Divergent MIC of Clarithromycin and Azithromycin Against Clinical Isolates of Mycobacterium abscessus Complex

Current treatment guidelines for Mycobacterium abscessus complex (MAB) infections recommend macrolides (clarithromycin/CAM and azithromycin/AZM) as cornerstone therapies, considering them clinically interchangeable without distinction. This study evaluated the comparative in vitro activity of clarithromycin (CAM) and azithromycin (AZM) against 146 clinical isolates of Mycobacterium abscessus complex organisms. MIC distributions revealed significantly higher resistance to AZM (94.5% resistant at [≥]8 g/mL) compared to CAM (77.4% resistant at [≥]8 g/mL), with only 4.8% of isolates susceptible to both drugs. Notably, 26 strains (17.8% of total) showed AZM resistance despite being non-resistant to CAM (MIC <8 g/mL) among 146 clinical isolates. Among the 33 CAM-non-resistant isolates (including susceptible and intermediate strains), 78.8% (26/33) were AZM-resistant, demonstrating frequent discordance in susceptibility profiles. These findings contrast with current treatment guidelines that recommend using CAM and AZM interchangeably without distinction. Subspecies analysis showed three distinct susceptibility profiles: isolates susceptible to both antibiotics (n=3, M. massiliense), isolates resistant to both antibiotics (n=3, M. massiliense), and M. bolletii with CAM-susceptible/AZM-resistant phenotypes (n=2). This differential resistance pattern suggests that while both drugs face substantial resistance challenges, CAM retains meaningful activity against certain strains that are AZM-resistant-particularly in M. bolletii. These findings indicate that AZM and CAM should not be used interchangeably to treat MAB infections and emphasize the need to prioritize CAM over AZM in treatment regimens.

microbiology↗

bioRxiv2025

Dynamic Pathway of Guanidine-III Riboswitch Folding Revealed by Single-Molecule FRET: Mg2+-Assisted Preorganization and Ligand-Induced Kinetic Trapping

Riboswitches are structured RNA elements that regulate gene expression by sensing and binding small molecules. The guanidine-III riboswitch, a critical bacterial regulator responding to guanidine toxicity, undergoes precise conformational changes that remain poorly characterized at a dynamic, mechanistic level. In this study, we employed single-molecule Forster Resonance Energy Transfer (smFRET) coupled with molecular dynamics (MD) simulations to delineate how the guanidine-III riboswitch transitions among distinct conformational states. We identify three principal states [-] an extended (E-state), a compacted prefolded intermediate (P-state), and a folded pseudoknot structure (F-state) [-] with rapid interconversion in the absence of ligand. Magnesium ions significantly stabilize intermediate states via a cooperative, preorganization mechanism, enhancing ligand binding efficiency. Binding of guanidine drastically suppresses the reverse transitions, kinetically trapping the riboswitch into its active folded state primarily through a conformational selection mechanism, with additional contributions from induced-fit dynamics. This work illuminates the unique dynamic pathway by which the guanidine-III riboswitch integrates ionic and ligand cues, ensuring precise gene regulatory responses in bacteria.

biophysics↗

bioRxiv2025

A humanized 16A antibody conjugated with DNA topoisomerase I inhibitors, targeting a GSTA glycosite-signature epitope.

BackgroundWe previously reported the 16A antibody, which binds to the abnormally glycosylated tandem repeat region of the MUC1 glycoprotein, and developed 16A-MMAE as an antibody-drug conjugate. However, its antitumor efficacy as an antibody-drug conjugate with DNA topoisomerase Iinhibitors remains unknown. MethodsWe humanized the 16A antibody and conjugated it to DXd and MF6, two DNA topoisomerase I inhibitors. The antitumor efficacy of the conjugates was evaluated in vitro and in vivo. ResultsThe humanized 16A-DXd conjugate showed potent antitumoral efficacy, with an IC50 in the nM range against CFPAC1 cells. In vivo, h16A-DXd and h16A-MF6 inhibited tumor growth in asubcutaneous mouse tumor transplantation model using CT26-COSMC KO-hMUC1 cells, administered at a dose of 10 mg/kg. ConclusionsThe high antitumor efficacy of h16A-conjugated DNA topoisomerase I inhibitors supports further clinical development. The relatively low toxicity of h16A-DXd and h16A-MF6 may enable a higher therapeutic window, since MUC1-positive cells internalize the antibody-drug conjugate more efficiently.

pharmacology and toxicology↗

bioRxiv2025

Multi-modal Disentanglement of Spatial Transcriptomics and Histopathology Imaging

Spatially-resolved expression profiling data has revolutionized biological research with multiple emerging clinical applications. Spatial transcriptomic assays are often jointly measured with histopathology imaging data, which is frequently used for diagnosing and staging various diseases. However, determining the extent to which the spatial transcriptomic and histopathology data represent overlapping or unique sources of variation is challenging, particularly given the myriad of factors influencing both, including expression variation, spatial context, tissue morphology, and batch effects. Here, we view this challenge as multi-modal disentanglement and develop an evaluation framework. We introduce SpatialDIVA, a disentanglement technique for jointly measured spatially resolved transcriptomics and histopathology data. We demonstrate that SpatialDIVA outperforms baseline techniques in disentangling salient factors of variation in curated pathologist-annotated multi-sample colorectal and pancreatic cancer cohorts. Further, SpatialDIVA removes batch effects from multi-modal data, allows for factor covariance analysis, and yields actionable biological insights through a novel conditional multi-modal generation method. The SpatialDIVA model, evaluation code, and datasets are available at https://github.com/hsmaan/SpatialDIVA.

genomics↗

bioRxiv2025

EpiFoundation: A Foundation Model for Single-Cell ATAC-seq via Peak-to-Gene Alignment

Foundation models exhibit strong capabilities for downstream tasks by learning generalized representations through self-supervised pre-training on large datasets. While several foundation models have been developed for single-cell RNA-seq (scRNA-seq) data, there is still a lack of models specifically tailored for single-cell ATAC-seq (scATAC-seq), which measures epigenetic information in individual cells. The principal challenge in developing such a model lies in the vast number of scATAC peaks and the significant sparsity of the data, which complicates the formulation of peak-to-peak correlations. To address this challenge, we introduce EpiFoundation, a foundation model for learning cell representations from the high-dimensional and sparse space of peaks. Epi-Foundation relies on an innovative cross-modality pre-training procedure with two key technical innovations. First, EpiFoundation exclusively processes the non-zero peak set, thereby enhancing the density of cell-specific information within the input data. Second, EpiFoundation utilizes dense gene expression information to supervise the pretraining process, aligning peak-to-gene correlations. EpiFoundation can handle various types of downstream tasks, including cell-type annotation, batch correction, and gene expression prediction. To train and validate EpiFoundation, we curated MiniAtlas, a dataset of 100,000+ single cells with paired scRNA-seq and scATAC-seq data, along with diverse test sets spanning various tissues and cell types for robust evaluation. EpiFoundation demonstrates state-of-the-art performance across multiple tissues and diverse downstream tasks.

genomics↗

bioRxiv2025

Non-Visual Light Sensing Enhances Behavioral Memory and Drives Gene Expression in C. elegans

Visible light influences a range of physiological processes, yet how animals respond to it independently of the visual system remains largely unknown. Here, we uncover a previously undescribed light-induced transcriptional pathway that modulates behavioral plasticity in C. elegans, a roundworm without eyes. We demonstrate that ambient visible light or controlled-intensity visible-spectrum LED activates an effector gene cyp-14A5 in non-neuronal tissues through the bZIP transcription factors ZIP-2 and CEBP-2. Light induction of cyp-14A5 is more prominent at shorter wavelengths but is independent of the known blue light receptors LITE-1 and GUR-3 in C. elegans. This bZIP-dependent genetic pathway in non-neuronal tissues enhances behavioral adaptability and olfactory memory, suggesting a body-brain communication axis. Furthermore, we use the light-responsive cyp-14A5 promoter to drive ectopic gene expression, causing synthetic light-induced sleep and rapid aging phenotypes in C. elegans. These findings advance our understanding of light-responsive mechanisms outside the visual system and offer a new genetic tool for visible light-inducible gene expression in non-neuronal tissues.

physiology↗

bioRxiv2024

Vispro improves imaging analysis for Visium spatial transcriptomics

Spatial transcriptomics (ST) enables the comprehensive analysis of gene expression while preserving the spatial context of tissues. The histological images accompanying ST data provide spatially cohesive information that is often challenging to capture through gene expression alone. However, analyzing such images is challenging due to the presence of fiducial markers and background regions, which can obscure important features and complicate downstream analysis. To address these challenges, we developed Vispro, an end-to-end, fully automated image processing tool tailored for 10x Visium data. Vispro integrates four key modules of fiducial marker detection, fiducial marker removal and image restoration, tissue region detection, and segmentation of disconnected tissue areas. We demonstrated that Vispro enhances the quality of ST images and improves the performance of downstream analyses, including tissue segmentation, cell segmentation, image registration, and histology-based gene imputation.

bioinformatics↗

bioRxiv2024

BrainScale: Enabling Scalable Online Learning in Spiking Neural Networks

Spiking neural networks (SNNs) represent a promising paradigm for understanding brain functions [1, 2] and developing neuromorphic intelligence [3, 4]. However, their potential remains largely unrealized due to a fundamental limitation: the lack of a scalable online learning system capable of supporting large-scale training of complex brain dynamics over long timescales. Current approaches are either limited to offline learning [5], impeding long-time task training; constrained by high memory complexity [6, 7], hindering network scaling; or using oversimplified models [8], failing to capture complex brain dynamics. Here, we introduce Brain-Scale, a system that synergizes model universality, computational efficiency, and engineering usability to enable scalable online learning in SNNs. First, BrainScale introduces standard model abstractions to support the training of diverse spiking networks. Second, BrainScale implements an online learning algorithm with linear memory complexity by exploiting intrinsic properties of SNN dynamics. Third, BrainScale provides an online learning compiler that automates the implementation of online learning for any user-defined models. Extensive evaluations across diverse SNN dynamics and computational tasks demonstrate that BrainScale consistently delivers strong training performance while maintaining extremely low memory usage and high computational efficiency. These properties enable the online training of a whole-brain-scale SNN of the Drosophila brain, accurately capturing functional activities across brain regions. These results position BrainScale as a foundational platform for advancing large-scale brain simulation and neuromorphic computing.

neuroscience↗

bioRxiv2024

Suppression of TGF-β/SMAD signaling by an inner nuclear membrane phosphatase complex

Cytokines of the TGF-{beta} superfamily control essential cell fate decisions via receptor regulated SMAD (R-SMAD) transcription factors. Ligand-induced R-SMAD phosphorylation in the cytosol triggers their activation and nuclear accumulation. We determined how R-SMADs are inactivated by dephosphorylation in the cell nucleus to counteract signaling by TGF-{beta} superfamily ligands. We showed that R-SMAD dephosphorylation is mediated by an inner nuclear membrane associated complex containing the scaffold protein MAN1 and the CTDNEP1-NEP1R1 phosphatase. Structural prediction, domain mapping and mutagenesis revealed that MAN1 binds independently to the CTDNEP1-NEP1R1 phosphatase and R-SMADs to promote their inactivation by dephosphorylation. Disruption of this complex led to nuclear accumulation of R-SMADs and aberrant signaling, even in the absence of TGF-{beta} ligands. These findings establish CTDNEP1-NEP1R1 as the elusive R-SMAD phosphatase and reveal the mechanistic basis for TGF-{beta} signaling inactivation and how this process is disrupted by disease-associated MAN1 mutations.

cell biology↗

bioRxiv2024

Trajectory-guided dimensionality reduction for multi-sample single-cell RNA-seq data reveals biologically relevant sample-level heterogeneity

The analysis of single-cell RNA-sequencing (scRNA-seq) data with multiple biological samples remains a pressing challenge. We present MUSTARD, a trajectory-guided dimension reduction method for multi-sample multi-condition scRNA-seq data. This all-in-one decomposition reveals major gene expression variation patterns along the trajectory and across multiple samples simultaneously, providing opportunities to discover sample endotypes along with associated genes and gene modules. In data-driven simulation, MUSTARD achieves high accuracy in distinguishing sample-level group differences that existing methods fail to capture. MUSTARD also demonstrates a robust ability to capture gene markers and pathways associated with phenotypes of interest across multiple real-world case studies.

bioinformatics↗