Cell Discovery

Heterogeneous nuclear ribonucleoprotein U (HNRNPU) deficiency is a rare genetic cause of neurodevelopmental disorders (NDDs) lacking targeted therapies. Here, we developed a transcriptomic-guided compound prioritization pipeline using Connectivity Map (CMap) analysis on multi-model transcriptomic signatures from HNRNPU-deficient human cells and mouse models. Ten compounds were selected through manual curation and functionally screened in patient-derived HNRNPU-deficient neuroepithelial stem (NES) cells with earlier observed cellular phenotypes. Two of the compounds, AS601245 and Lenalidomide, significantly reduced the elevated neural progenitor population during differentiation, and their combination further decreased primary cilia incidence, indicating partial rescue of the patient-specific cellular phenotypes. To understand the mechanisms underlying the partial rescue, we employed proteome integral solubility alteration (PISA) and expression proteomics. PISA assay identified TMEM150C and GSK3A as proximal targets of combined treatment. Additionally, we observed reversal of multiple biological pathways including downregulation of Wnt signalling and upregulation of mitochondrial pathways and transmembrane proteins. Altogether, we established a computational-experimental pipeline for transcriptomic-guided drug repurposing for a monogenic NDD, and demonstrated that the network-level modulation partially rescues the delayed neural differentiation in HNRNPU-deficient neural cells.

Bifidobacteriaceae and Lactobacillaceae are key probiotic families and widely used in food production, yet a comprehensive understanding of strain functions and their gut microbial interactions based on complete genomes remain understudied. Here we constructed a complete-genome dataset of 3,300 strains from these two families, including 1,151 newly isolated from China. Compared with draft assemblies, complete genomes substantially recovered a gene functional landscape encompassing stress tolerance, surface exopolysaccharide synthesis, nutrient utilization, and mobile genetic elements. Major species from both families exhibited a prevalence >60% in the Chinese population, far higher than that in US/Dutch cohorts. Notably, as a core probiotic species with remarkable genomic plasticity and gut-adaptive potential, Lactiplantibacillus plantarum stood out in our dataset for its enriched functional profile and was particularly abundant in the Chinese population. Moreover, compared with non-Chinese genomes, our isolates of key species displayed less metabolic complementarity and stronger competition with potentially pathogenic keystone species in the gut, thereby linking strain origin to enhanced probiotic potential and ecological fitness to benefit human gut health.

Human immunodeficiency virus (HIV) infection remains a global health threat. Although antiretroviral therapy (ART) has significantly transformed HIV into a manageable chronic disease, emergence of drug resistance to current ART is a continuing concern. Broadly neutralizing antibodies, as well as reagents containing both a soluble CD4 mimetic and an HIV co-receptor inhibitor--such as CD4-CD4i antibodies--are promising strategies for the prevention and treatment of HIV infection. We previously developed a CD4 D1mimetic (mD1.22) with enhanced neutralization potency, surpassing that of the clinically validated sCD4-D1D2 mimetic. We also previously identified a novel CD4i antibody (X5) that targets the conserved coreceptor binding site on the gp120 core and recognizes an epitope partially overlapping with 17b monoclonal antibody binding site. X5 binding to gp120 was augmented by CD4 and modestly enhanced by CCR5. To leverage these favorable interactions, we designed and optimized sCD4-X5 bispecific antibodies by computational structure-aided modification of antibody size and fusion linkers between the two binding moieties. The bispecific D1X5, with a (G4S)7 long linker between D1 and X5 (IgG1-LL D1X5), exhibited broad neutralization against 11 diverse HIV subtypes across B, C, G clades and AC, BC recombinants. The TZM-bl cell neutralization assay showed IgG1-LL D1X5 neutralization geometric mean IC50 and IC80 are 0.6 g/mL and 3.4 g/mL respectively, which are within the range of potent bnAbs. This work has identified a novel single domain soluble CD4 based CD4-CD4i bispecific antibody with broad HIV-1 neutralization.

Acinetobacter baumannii is a top-priority, ESKAPE pathogen that poses a major challenge to human health. The pathogen is difficult to combat due to its extensive arsenal of antibiotic resistance and its protective polysaccharide capsule. In addition, A. baumannii isolates are highly heterogeneous, which complicates the development of rapid detection methods or novel targeted therapeutic approaches. Here, we discovered and characterized a new biotechnological tool, the nanobody H7 (NbH7), along with its conserved target, the surface-exposed Omp25 protein of A. baumannii, and elucidated their interaction at the molecular level. Moreover, we demonstrate that NbH7-functionalized magnetic beads enable selective and efficient capture of A. baumannii from bacterial mixtures, including non-pathogenic intestinal bacteria. This provides proof of concept for a new targeting system that remains effective across diverse A. baumannii clinical isolates and capsule types and holds potential for use in diagnostic cell enrichment and targeted therapies.

The SARS-CoV-2 main protease (Mpro) has been a major target of antiviral drug development, leading to the development of inhibitors such as nirmatrelvir, the antiviral component of the COVID-19 drug Paxlovid. However, resistance-associated mutations that reduce the efficacy of current Mpro inhibitors, particularly nirmatrelvir, have emerged. Here, we evaluated the inhibitory activity of leritrelvir (RAY1216), an Mpro inhibitor approved in China for COVID-19 monotherapy, against a panel of Mpro variants carrying mutations at 12 resistance-associated residues distributed across four catalytic subsites. Using integrated biochemical, biophysical, structural, and cellular analyses, we demonstrate that leritrelvir retains stronger inhibitory activity against most tested resistant mutants compared with nirmatrelvir. Most of the tested mutations promote Mpro dimer dissociation, with E166V showing a particularly pronounced effect and markedly compromising nirmatrelvir binding. In contrast, thermal shift and size-exclusion chromatography assays demonstrate that leritrelvir binding restores dimerization of these Mpro mutants. Sixteen high-resolution crystal structures reveal that leritrelvir binding re-instates key dimer-interface interactions disrupted by resistance mutations. Mini-replicon assays further confirm leritrelvir to possess enhanced cellular antiviral efficacy compared with nirmatrelvir. Our findings indicate that tighter leritrelvir binding enables more effective inhibition of dissociation-prone Mpro mutants than nirmatrelvir, supporting its use as a more resilient antiviral agent for SARS-CoV-2 treatment.

As the recent COVID-19 pandemic illustrated, zoonotic viruses and other pathogens pose a credible threat to public health. Recent advancements in vaccine technology, particularly mRNA vaccines, provide key tools for an effective and swift public health response. Although mRNA vaccines can be developed more quickly than traditional vaccines, fast and accurate construction of DNA templates for these vaccines remains a critical bottleneck. Using our novel DNA assembly technology, Sidewinder, we rapidly designed and built multiple mRNA vaccine candidates to guard against a potential outbreak of Hantavirus (ANDV). We successfully constructed the DNA templates from oligo pools and produced the mRNA for three vaccine candidates in just 2 days after delivery of the synthetic DNA oligos.

Human genetic studies have identified defects in multiple mechanisms that predispose the risk of developing inflammatory bowel diseases (IBD), which include alterations in adaptive and innate immune responses, epithelial integrity and regulation of the intestinal mucus layer. Despite the importance of intestinal barrier integrity in the pathogenesis of IBD, essentially all current therapies modulate the immune responses. In this study, we determined the high resolution cryo-EM structure of human NXPE1, a IBD associated protein. Based on the structural homology, we identified NXPE1 as an O-acetyltransferase. Since NXPE1 is a pseudo gene in mouse, we generated knockout mouse model that lacked two of the mouse NXPE1 homologs, Nxpe2 and Nxpe4. The O-acetylation of sialic acid on red blood cells was abolished in the double knockout mice, confirming the sialic acid O-acetyltransferase function of NXPE1 family members. These findings underscore the potential of NXPE1 as a novel therapeutic target of the intestinal barrier functions for the treatment of IBD.

The developmental program governing meibomian gland (MG) morphogenesis and proliferation remains poorly understood, largely due to the lack of physiologically relevant model systems. Here, we established a novel high-fidelity, three-dimensional organoids model derived from mouse meibomian gland (mMGO) epithelium. Transcriptomic and phenotypic analyses demonstrated that mMGOs faithfully recapitulate postnatal gland development in vivo, including dynamic transcription program, branching morphogenesis, lineage differentiation, and functional lipid accumulation. Leveraging this model, we identified the Hippo-YAP pathway as a pivotal regulator of MG epithelial proliferation and homeostasis for the first time. YAP inhibition severely impaired organoids growth, while pharmacological inhibition of Hippo pathway with XMU-MP-1 enhanced proliferation and progenitor clonogenicity. Crucially, in inflammation-induced atrophic organoids, XMU-MP-1 treatment rescued YAP nuclear localization and stimulated regrowth and functional restoration. Our study provided new mechanistic insights and a robust organoids platform for MG development research, and nominated targeted Hippo pathway inhibition as a promising strategy to reverse glandular atrophy in meibomian gland dysfunction.

Despite decades of progress in computational drug discovery, deep learning-based molecular representation models remain largely structure-centric, assuming that chemical similarity approximates functional similarity. However, drug effects in cells are shaped not only by chemical similarity but also by molecular interactions in the cellular context. To capture this complexity, we introduce CellChem, a cellular-chemical representation-learning framework that learns cell-guided molecular representations of small-molecule actions within cells. By incorporating large-scale cellular transcriptional profiles during pretraining, CellChem reshapes molecular representation space from structure-centric to a balanced integration of structure and function. The learned CellChem molecular representations exhibit biologically meaningful geometric organization, such that distances between molecules encode not only structural similarity but also similarity in the cellular responses they elicit, independent of downstream tasks. Using downstream derivative models such as cell-guided compound-protein interaction prediction and drug-induced transcriptional response profile generation, CellChem supports highly efficient, multi-scale drug discovery, achieving significantly better performance than traditional models.

Bacteria have evolved multiple immune systems to resist phage invasion, however, only a small part of the defensive mechanisms have been clearly uncovered. In this study, we report a type III Druantia two-component defense system, DruH-DruE, identified from Mycobacterium smegmatis. The DruH-DruE prevents phage DNA cyclization and replication.DruE can be replaced from the defense system by either homolog in M. tuberculosis or M. smegmatis. The physical interaction between this two components is essential for fighting against phage infection. Mutations in the interaction sites led to the loss of phage-defending function of the system. The broad-spectrum antiphage ability of the defense system could be activated by the small tail protein Gp25 of phage A10ZJ24. This study fills a major gap in current knowledge of antiphage mechanism of type III Druantia defense system, expanding our understanding of the immune mechanisms in prokaryotic cells.

Translation efficiency remains a major limitation for RNA therapeutics. Conventional optimization targets the 5 untranslated region (5 UTR), while the 3 UTR is viewed mainly as a stabilizing element. Here, we demonstrate that the 3 UTR can be rationally engineered to actively enhance translation. Using an intracellular directed-evolution platform based on the SINEB2 element, we identified RNA modules P51 and its compact variant P51t3,which markedly increased protein output without affecting mRNA levels. P51t3 consistently boosted expression two- to six-fold across plasmid, in vitro transcribed mRNA, and recombinant AAV systems. Mechanistic studies revealed that P51t3 binds ribosomal protein RPL39, recruiting 60S subunits to the initiation site through the natural closed-loop translation model. By integrating evolutionary selection with 3 UTR design, this work redefines the 3 UTR as an active translational enhancer and provides a broadly applicable regulatory element for next-generation mRNA and gene-delivery therapeutics.

Calreticulin is an emerging cancer biomarker, but current detection methods rely on expensive monoclonal antibodies that suffer from inefficient protein production, pharmacokinetic challenges and poor tissue penetration. Cal3, a calreticulin-specific nanobody, was constructed by replacing the complimentary determining region 2 (CDR2) of a soluble, clinically validated nanobody with a calreticulin-specific CDR2 isolated from a phage display library. However, the poor solubility and low yield of Cal3 limit its usefulness. In this study, we engineered CALR-Nb02 by adapting the core of Cal3 to a partial consensus framework sequence of stable nanobodies. CALR-Nb02 was purified with a 240-fold higher yield as a predominantly monomeric, soluble protein that exhibits an increased thermal stability and a higher calreticulin binding affinity (KD: 25-50 nM) compared with Cal3. These results reveal a strategy for quickly altering the specificity of a stable nanobody, and provide an improved calreticulin-binding reagent for future diagnostic, imaging, and therapeutic applications.

Olfaction enables animals to detect and discriminate a vast array of chemicals, yet how odorant receptors (ORs) encode ligand selectivity remains unclear. Although recent advances in protein structure prediction have expanded access to OR structures, linking these to function at scale has been challenging. Here, we combined AlphaFold3-predicted receptor structures with protein language model embeddings and in vivo pS6-IP-Seq measurements of olfactory sensory neuron activation across a chemically diverse odor panel to train a deep learning model of OR-ligand interactions. The resulting framework predicts receptor responses, organizes receptors by functional similarity independent of sequence, and identifies structural determinants of ligand selectivity. These findings establish a structure-based map of OR function and provide a foundation for predictive and interpretable models of olfactory coding.

Vaccination remains the most successful preventative measure against viral infection, but methods to stably deter rapidly-evolving pathogens have remained elusive. Vaccines capable of incorporating and anticipating viral evolution could address current challenges in seasonal vaccination efforts against SARS-CoV-2 and influenza where economic and disease burdens remain high despite decades of combined study. Rare epitope suppression (RES) is an underutilized concept within vaccine design, where humoral epitope targeting can be molded using complex antigen pools. Based in mRNA vaccine technology, 'wobble vaccines' represent the novel application of RES to human pathogens designed to anticipate and resist viral evolution. To establish this platform, public SARS-CoV-2 sequencing data was compiled from the first two years of the COVID-19 pandemic to identify high-diversity sites across the receptor binding domain (RBD) of the spike protein. Wobble RBD (WobbRBD) libraries reflecting that entropy were synthesized and incorporated into established self-amplifying (SA) vaccine constructs. Animals immunized with these complex antigen pools showed no obvious adverse effects. By three days-post vaccination, WobbRBD stimulated robust primary immune activation with distinctive characteristics compared to traditional single-strain vaccine modalities. By day 14, germinal centers, class switching, and antibody-secreting cells were induced, creating potent SARS-CoV-2 spike-binding IgG antibodies. Despite similar overall activation profiles, WobbRBD generated significantly increased breadth against SARS-CoV-2 variant spikes in comparison to single-strain controls -- even against future-emerging strains. Taken together, wobble vaccines represent a novel method for anticipating and preventing viral escape with promising applications in SARS-CoV-2, influenza, HIV, and beyond.

Glycosyltransferases often contain multiple structural modules that contribute to substrate recognition, catalytic coordination, and higher-order molecular organization. However, how multidomain glycosyltransferases dynamically organize their catalytic domains in solution remains poorly understood. Here, we investigated the assembly states and conformational dynamics of POMGNT2, LARGE1, K4CP, and L137 using high-speed atomic force microscopy (HS-AFM) integrated with complementary solution biophysical analyses. POMGNT2 formed a stable dimeric architecture with limited large-scale conformational fluctuation, consistent with its role in site-selective substrate recognition. In contrast, LARGE1 and K4CP exhibited concentration-dependent and heterogeneous assembly behavior. K4CP displayed pronounced open-closed interdomain motion and substrate-dependent conformational compaction, indicating dynamic catalytic-domain reorganization during glycan elongation. By comparison, the mimivirus glycosyltransferase candidate L137 predominantly behaved as a monomeric species under the tested conditions. These findings demonstrate that multidomain glycosyltransferases employ diverse dynamic organizational strategies ranging from rigid recognition architectures to highly flexible and reversible catalytic assemblies. Our results further suggest that glycosyltransferase function is governed not only by catalytic-domain structure, but also by dynamic conformational coordination adapted to distinct catalytic demands.

Extreme environments, though hostile to most life forms, host specialized extremophile communities that have redefined biological cognition and emerged as vital biotechnological resources, with their unique adaptive traits and bioactive molecules driving advances in multiple scientific and industrial fields. However, research on extremophiles is hindered by limitations in culture-based methods, fragmented multi-omics data with non-uniform annotation standards across repositories, the lack of cross-extreme comparative research in existing resources, and the singularity of data dimensionality that neglects key structural information, all of which restrict the functional interpretation of extremophile microbes and the exploitation of their bioprospecting potential. To tackle these challenges, we developed ExMODE (https://db.genomics.cn/exmode/), a comprehensive multi-omics database platform dedicated to extremophiles. It centrally integrates multi-omics data from diverse extreme habitats with a standardized annotation framework, resolving data fragmentation and enabling systematic cross-environment comparative analyses to elucidate extremophile adaptive mechanisms. Moreover, ExMODE aggregates multi-dimensional datasets including genes, genomes, secondary metabolite sequences and protein structures, overcoming the constraints of single-dimensional data and significantly improving the efficiency of biotechnological resource discovery from extreme microorganisms.

The integration of in vitro selection, deep sequencing, and machine learning (ML) has recently been developed as a powerful strategy for discovering functional antibodies. However, how training data composition and ML search space design influence the identification of high-affinity variants remains unclear. Here, we aimed to optimize ML-integrated directed evolution for functional antibody discovery by selecting training data from deep sequencing analysis. By performing phage display selection using camelid heavy-chain antibodies (VHHs), we demonstrated that early-round data, retaining more binding-negative variants, can be superior for training models to identify high-performance VHHs. We also investigated a lead-independent ML search space design by focusing on conserved residues in final rounds, successfully identifying variants with higher affinities than those from lead-based maturation (KD = 7.9 nM). These findings demonstrate that training data selection and search space design are critical for successful ML-guided antibody engineering and provide diverse pathways for discovering high-affinity VHH variants.

The Rcs phosphorelay regulates gene expression in response to cell envelope stress and is critical for the virulence of pathogenic bacteria, including Klebsiella pneumoniae, due to its regulation of genes related to extracellular capsule, cell division, and motility. The RcsC histidine kinase, RcsD phosphotransfer protein and RcsB response regulator, which form the core of the Rcs phosphorelay, are negatively regulated by the unique inner membrane protein IgaA via interaction with RcsD. An outer membrane lipoprotein, RcsF, activates signaling by interaction with IgaA, but the precise activation mechanisms remain unclear. In this study, we determined the structures of IgaA and the IgaA/RcsF complex using Cryo-electron microscopy (Cryo-EM). We also determined the structures of RcsC and RcsD, which both form homodimers stabilized by hydrophobic interactions, creating ladder-like structures. Combining the Cryo-EM structures, AlphaFold3 structure predictions of IgaA/RcsD and RcsF/IgaA/RcsD, and genetic studies, we describe a model for how RcsF modifies the IgaA/RcsD interaction, lifting negative regulation and activating the Rcs phosphorelay. Our findings provide a high-resolution depiction of the Rcs stress response system and suggest potential targets for small molecule inhibitors.

The Nucleosome Remodeling and Deacetylase complex (NuRD) plays a key role in regulating hemoglobin expression in adult erythroid cells. Selectively disrupting this complex potently induces the expression of fetal hemoglobin, a proven therapeutic strategy for treating beta-hemoglobinopathies such as sickle cell anemia. In these studies, we have used mRNA display to identify small macrocyclic peptides that inhibit the interaction between two core components of NuRD, the SANT-SLIDE domain of CHD4 and the CR2 domain of GATAD2A. In addition, the screen suggested a second binding site on the CHD4 domain. Based on this observation, we hypothesized and confirmed that CDK2AP1 bound to this region of CHD4, leading us to purify and determine the structure of the ternary complex between CHD4, GATAD2A, and CDK2AP1. The results of our studies show that the SANT-SLIDE domain of CHD4 functions as a critical interaction hub in the formation of NuRD and suggest a strategy to block NuRD function for therapy.

Conventional structure-based drug design has high clinical failure rates due to the disconnect where binding affinity does not guarantee safe functional modulation. To bridge this gap, we present CRISPRtile, a cloud-based platform for function-based drug design. By deriving library coverage optimization equations and leveraging AI to correct CRISPR guide biases, we generated toxicity and functional landscapes with over threefold error reduction compared to conventional methods. These maps bypass error and orders of magnitude higher computational cost in structure-based pipelines by enabling AI prediction of drug interaction directly from nontoxic functional sequences, while predicting brain penetration with benchmark leading performance. We demonstrate CRISPRtile by mapping the NLRP3 inflammasome and identifying FDA approved drugs with previously unrecognized ability to modulate it, revealing strategies to amplify or inhibit our immune response to homeostatic perturbations. These advances establish a generalizable strategy for the systematic discovery of safe functional modulators.