Molecular Systems Biology最新文献

Deep Mutational Scanning (DMS) assays can systematically assess the effects of amino acid substitutions on protein function, but many datasets have incomplete variant coverage due to technical constraints. We developed VEFill (Variant Effect Fill), a gradient boosting model for imputing missing DMS scores across protein domains. Trained on the Human Domainome 1, VEFill integrates ESM-1v sequence embeddings, evolutionary conservation (EVE scores), amino acid substitution matrices, and physicochemical descriptors. The model achieved robust predictive performance (Pearson r = 0.80) and generalized reliably to unseen proteins in stability-based datasets, while showing weaker performance on activity-based assays. Per-protein models confirmed VEFill's effectiveness under limited-data conditions and a reduced two-feature version performed comparably to the full model, suggesting an efficient alternative. Across multiple benchmarking settings, VEFill consistently outperformed baselines once ≥20% of experimental measurements were available. However, true zero-shot prediction without positional context remains challenging, particularly for functionally complex proteins. Overall, VEFill offers an interpretable, scalable framework for DMS score imputation, and enables systematic mutation prioritization including the design of sparse experimental libraries for variant effect studies.

Viral proteins interact with host proteins to hijack cellular pathways important for viral replication. Viral mimics are proteins whose structural similarity to host-mimicked proteins allows them to interact with mutual host targets. This mimicry poses a challenge for the host-how to avoid mimics without compromising essential interactions with host-mimicked proteins. Despite the prevalence of mimicry, the evolutionary dynamics between host and viral mimics remain largely unknown. We address this by integrating structural modeling, host-virus interaction networks, and comprehensive evolutionary analyses of host and viral proteins. We show that host proteins targeted by mimics and host-mimicked proteins are highly conserved, and that this is related to functional constraints imposed on host proteins. Host interface residues that interact with both mimics and host-mimicked proteins evolve slowly, while residues that exclusively interact with mimics evolve significantly faster. Surprisingly, viral mimics do not evolve rapidly, instead displaying complex evolutionary patterns. Our analysis reveals host's limited capacity to escape mimicry and viral evolution to exploit this, and highlights how constraints lead to unexpectedly slow evolution of host-virus interaction networks.

Microbial surface functionalization is a powerful strategy for endowing microbes with novel, non-genetic functions. However, existing methods are often species-specific, limited in scope, and compromise cell viability. Here, we present a universal and modular platform for high-density, reproducible surface functionalization across diverse microbial species-including Gram-positive, Gram-negative, aerobic, and anaerobic bacteria-using multiple molecular classes such as fluorophores, enzymes, and nucleic acids. Our method preserves cell viability and achieves 50× higher functionalization efficiency than previous methods with a standardized protocol applicable to any azide-containing molecule. Applications of the method show reproducible and tunable phenotypic outcomes at the single-cell level: fluorophore labeling yielded adjustable fluorescence, β-lactamase conferred scalable antibiotic resistance, and DNA coatings modulated adhesion and aggregation. This platform provides quantitative, non-genetic control over microbial phenotypes and complements genetic engineering approaches. It enables new possibilities for microbial design in biotechnology, medicine, and environmental applications where genetic modification is impractical or undesirable.

The aggregation of the protein alpha-synuclein (αSyn) is a common feature of multiple neurodegenerative diseases collectively called synucleinopathies, for which the pathobiology is not well understood. The different phenotypic characteristics of the synucleinopathies Parkinson's disease (PD), Dementia with Lewy Bodies (DLB) and Multiple System Atrophy (MSA) have been proposed to originate from the distinct structures adopted by αSyn in its amyloid forms. Here, using covalent labeling and limited proteolysis coupled to mass spectrometry (LiP-MS) in vitro and in situ within neuronal cells and directly in native patient brain homogenates, we show that pathogenic αSyn from distinct synucleinopathies (PD, DLB and MSA) are structurally different. Further, we found that fibril structural differences are associated with different putative fibril interactomes and neuronal responses. We discovered disease-specific ubiquitination patterns and turnover profiles for pathogenic αSyn species, detected molecular pathways responding specifically to the uptake of different αSyn fibrillar polymorphs, and identified a subset of the involved proteins as putative interactors of αSyn. In particular, components of the ubiquitin-proteasomal System (UPS), including E3 ubiquitin ligases, chaperones, and deubiquitinating proteins, showed disease/polymorph-specific putative interaction patterns, possibly accounting for different resistance of patient-derived αSyn fibrils to degradation. Genetic modulation with CRISPR-based tools showed that members of the UPS degradation pathway (three E3 ligases: UBE3A, TRIM25, HUWE1 and the AAA+ ATPase VCP) reduced αSyn inclusions, in a strain-specific manner. LiP-MS also identified sets of proteins with altered protease susceptibility in postmortem brain homogenates of PD, DLB, and MSA patients. These sets were largely disease-specific and included proteins altered in cells treated with fibrils derived from patients with the matching disease. Our findings provide insight into cellular processes involved in the accumulation and turnover of αSyn pathogenic aggregates in PD, DLB and MSA in a disease/specific manner and constitutes a resource of potential novel drug targets in these synucleinopathies.

Tumor microenvironment (TME) is characterized by a milieu of distinct cell types that exist in heterogenous transcriptional states across tumors. Functional interactions among these cell states drive tumor progression and therapy response. Systematic characterization of functional cell-state interactions (CSIs) remains challenging due to the paucity of scRNA-seq cohorts with clinical information on one hand, and the lack of cellular context in bulk RNA-seq cohorts on the other. We present CSI-TME, a computational pipeline that extends the concept of gene interactions, such as synthetic lethality, to cell states, to infer prognostic CSIs by directly leveraging large cohorts of bulk transcriptomic datasets. Applied CSI-TME to IDH-mutant gliomas, we identified a highly reproducible cell-state interaction network (CSIN) that is predominantly pro-tumor and differentially activated in IDH-mut astrocytoma versus oligodendroglioma. Malignant cell states within the CSIN resemble multiple neuronal lineages, including astrocyte-like and oligodendrocyte-progenitor-like programs, and reveal key interactions between glioma stem cells and T cells. CSIN stratifies patient response to immune-checkpoint blockade therapy. Roughly 20% of CSIs involve direct ligand-receptor communication, and co-localize in spatial-transcriptomic datasets, most notably for a pro-tumorigenic interaction between tip-like endothelial cells and hypoxic malignant cells supported by multiple ligand-receptor interactions. Interestingly, anti-tumor CSIs correlated with oncogenic mutations are preferentially active in early stages of cancer, hinting at tissue homeostatic response. Overall, CSI-TME is a novel approach that, leveraging clinical bulk transcriptomic data, identifies prognostic CSIs and therapeutic ligand-receptor targets, while providing novel insight into how interactions among the cell states shape the TME in IDH-mutant glioma.

Oncogenic mutations shape colorectal cancer (CRC) biology, yet their impact on transcriptional phenotypes remains incompletely understood, and their individual prognostic value is limited. Here, we perform a pooled single-cell transcriptomic screen of over 100,000 CRC cells with a comprehensive barcoded library of oncogenic variants across genetically diverse CRC lines. Using a variational autoencoder-based interpretable factor model, we identify ten conserved oncogene-driven transcriptional modules (TMOs) representing core cancer phenotypes such as cellular plasticity, inflammatory response, replicative stress, and epithelial-to-mesenchymal transition. Engagement of these modules can be context-dependent, reflecting interactions between oncogene-induced driver pathways and background genetics. TMO activity in patient tumors stratifies CRC cohorts into high- and low-risk groups, improving relapse-free survival prediction beyond existing classification systems. Our study systematically links oncogenic signaling to transcriptional states and clinical outcomes, establishing a functional framework for module-based patient stratification in precision oncology.

Single-cell RNA sequencing allows defining cellular identities based on transcriptional similarity using unsupervised clustering. However, a single clustering resolution may not yield groups of cells that represent both broad, well-defined populations and smaller subpopulations simultaneously. Therefore, when cell identities are not known prior to sequencing, robust comparison and annotation of inferred de novo clusters remains a challenge. Here, we introduce CANTAO, in which we propose the average overlap metric to define the distance between single-cell clusters by comparing ranked lists of differentially expressed genes in a top-weighted manner. We benchmark CANTAO in truth-known datasets comprised of similar yet distinct cell populations and show that evaluating clusters with average overlap results in a consistent, precise, and biologically meaningful recapitulation of true cell identities. We then analyze unsorted mouse thymocytes and characterize stages of T-cell development in the thymus, including minor populations of double-negative (CD4-CD8-) T cells that are difficult to confidently detect among unsorted single cells. We demonstrate that CANTAO enables robust, reproducible characterization of single-cell data and clarifies biological interpretation of underlying identities in homogeneous populations.

Persisters represent a transient, antibiotic-tolerant subpopulation within isogenic bacterial populations, contributing to infection relapses. However, the mechanisms driving persister formation and resuscitation remain elusive. Here, we developed nano-flow cytometry (nFCM)-based methods for single-cell quantification of toxin (T) RelE and antitoxin (A) RelB levels, as well as for monitoring persister states through cell wall growth. We demonstrate that bacteria elevate the T/A ratio through two distinct TA expression modalities to withstand bacteriostatic antibiotic challenge, with T/A = 1.0 as a critical threshold. Intriguingly, single-cell resuscitation dynamics revealed that subinhibitory antibiotic exposure promotes entry into a deeper dormant state characterized by elevated T/A ratios, underscoring the importance of maximizing therapeutic antibiotic concentrations. Crucially, we uncovered a triphasic detoxification process during resuscitation where progressive toxin depletion drives T/A ratio reduction to a critical proliferation-permissive threshold. Proteomic profiling unveiled that persisters with high RelE production have increased transmembrane transporter levels linked to stress response and drug efflux. Our findings offer pivotal molecular insights underlying persister transitions and underscore the need for high-throughput, single-cell analysis of these heterogeneity phenotypes.

Bacterial colonization of tumors is widespread, yet the dynamics during colonization remain underexplored. Here we discover strong variability in the sizes of intratumor bacterial clones and use this variability to infer the mechanisms of colonization. We monitored bacterial population dynamics in murine tumors after introducing millions of genetically barcoded Escherichia coli cells. Results from intravenous injection revealed that roughly a hundred bacteria seeded a tumor and that colonizers underwent rapid, yet highly nonuniform growth. Within a day, bacteria reached a steady-state and then sustained load and clone diversity. Intratumor injections, circumventing colonization bottlenecks, revealed that the nonuniformity persists and that the sizes of bacterial progenies followed a scale-free distribution. Theory suggested that our observations are compatible with a growth model constrained by a local niche load, global resource competition, and noise. Our work provides the first dynamical model of tumor colonization and may allow distinguishing genuine tumor microbiomes from contamination.