We envision "AI scientists" as systems capable of skeptical learning and reasoning that empower biomedical research through collaborative agents that integrate AI models and biomedical tools with experimental platforms. Rather than taking humans out of the discovery process, biomedical AI agents combine human creativity and expertise with AI's ability to analyze large datasets, navigate hypothesis spaces, and execute repetitive tasks. AI agents are poised to be proficient in various tasks, planning discovery workflows and performing self-assessment to identify and mitigate gaps in their knowledge. These agents use large language models and generative models to feature structured memory for continual learning and use machine learning tools to incorporate scientific knowledge, biological principles, and theories. AI agents can impact areas ranging from virtual cell simulation, programmable control of phenotypes, and the design of cellular circuits to developing new therapies.

Here, we describe "obelisks," a class of heritable RNA elements sharing several properties: (1) apparently circular RNA ∼1 kb genome assemblies, (2) predicted rod-like genome-wide secondary structures, and (3) open reading frames encoding a novel "Oblin" protein superfamily. A subset of obelisks includes a variant hammerhead self-cleaving ribozyme. Obelisks form their own phylogenetic group without detectable similarity to known biological agents. Surveying globally, we identified 29,959 distinct obelisks (clustered at 90% sequence identity) from diverse ecological niches. Obelisks are prevalent in human microbiomes, with detection in ∼7% (29/440) and ∼50% (17/32) of queried stool and oral metatranscriptomes, respectively. We establish Streptococcus sanguinis as a cellular host of a specific obelisk and find that this obelisk's maintenance is not essential for bacterial growth. Our observations identify obelisks as a class of diverse RNAs of yet-to-be-determined impact that have colonized and gone unnoticed in human and global microbiomes.

Ovarian cancer is resistant to immunotherapy, and this is influenced by the immunosuppressed tumor microenvironment (TME) dominated by macrophages. Resistance is also affected by intratumoral heterogeneity, whose development is poorly understood. To identify regulators of ovarian cancer immunity, we employed a spatial functional genomics screen (Perturb-map), focused on receptor/ligands hypothesized to be involved in tumor-macrophage communication. Perturb-map recapitulated tumor heterogeneity and revealed that interleukin-4 (IL-4) promotes resistance to anti-PD-1. We find ovarian cancer cells are the key source of IL-4, which directs the formation of an immunosuppressive TME via macrophage control. IL-4 loss was not compensated by nearby IL-4-expressing clones, revealing short-range regulation of TME composition dictating tumor evolution. Our studies show heterogeneous TMEs can emerge from localized altered expression of cancer-derived cytokines/chemokines that establish immune-rich and immune-excluded neighborhoods, which drive clone selection and immunotherapy resistance. They also demonstrate the potential of targeting IL-4 signaling to enhance ovarian cancer response to immunotherapy.

Brain recordings collected at different resolutions support distinct signatures of neural coding, leading to scale-dependent theories of brain function. Here, we show that these disparate signatures emerge from a heavy-tailed, multiscale functional organization of neuronal activity observed across calcium-imaging recordings collected from the whole brains of zebrafish and C. elegans as well as from sensory regions in Drosophila, mice, and macaques. Network simulations demonstrate that this conserved hierarchical structure enhances information processing. Finally, we find that this organization is maintained despite significant cross-scale reconfiguration of cellular coordination during behavior. Our findings suggest that this nonlinear organization of neuronal activity is a universal principle conserved for its ability to adaptively link behavior to neural dynamics across multiple spatiotemporal scales while balancing functional resiliency and information processing efficiency.

The autophagy-lysosome system directs the degradation of a wide variety of cargo and is also involved in tumor progression. Here, we show that the immunity-related GTPase family Q protein (IRGQ), an uncharacterized protein to date, acts in the quality control of major histocompatibility complex class I (MHC class I) molecules. IRGQ directs misfolded MHC class I toward lysosomal degradation through its binding mode to GABARAPL2 and LC3B. In the absence of IRGQ, free MHC class I heavy chains do not only accumulate in the cell but are also transported to the cell surface, thereby promoting an immune response. Mice and human patients suffering from hepatocellular carcinoma show improved survival rates with reduced IRGQ levels due to increased reactivity of CD8+ T cells toward IRGQ knockout tumor cells. Thus, we reveal IRGQ as a regulator of MHC class I quality control, mediating tumor immune evasion.

Chronic itch is a debilitating symptom profoundly impacting the quality of life in patients with liver diseases like cholestasis. Activation of the human G-protein coupled receptor, MRGPRX4 (hX4), by bile acids (BAs) is implicated in promoting cholestasis itch. However, the detailed underlying mechanisms remain elusive. Here, we identified 3-sulfated BAs that are elevated in cholestatic patients with itch symptoms. We solved the cryo-EM structure of hX4-Gq in a complex with 3-phosphated deoxycholic acid (DCA-3P), a mimic of the endogenous 3-sulfated deoxycholic acid (DCA-3S). This structure revealed an unprecedented ligand-binding pocket in MRGPR family proteins, highlighting the crucial role of the 3-hydroxyl (3-OH) group on BAs in activating hX4. Guided by this structural information, we designed and developed compound 7 (C7), a BA derivative lacking the 3-OH. Notably, C7 effectively alleviates hepatic injury and fibrosis in liver disease models while significantly mitigating the itch side effects.

Genetic variation in host immunity impacts the disproportionate burden of infectious diseases that can be experienced by First Nations peoples. Polymorphic human leukocyte antigen (HLA) class I and killer cell immunoglobulin-like receptors (KIRs) are key regulators of natural killer (NK) cells, which mediate early infection control. How this variation impacts their responses across populations is unclear. We show that HLA-A∗24:02 became the dominant ligand for inhibitory KIR3DL1 in First Nations peoples across Oceania, through positive natural selection. We identify KIR3DL1∗114, widespread across and unique to Oceania, as an allele lineage derived from archaic humans. KIR3DL1∗114+NK cells from First Nations Australian donors are inhibited through binding HLA-A∗24:02. The KIR3DL1∗114 lineage is defined by phenylalanine at residue 166. Structural and binding studies show phenylalanine 166 forms multiple unique contacts with HLA-peptide complexes, increasing both affinity and specificity. Accordingly, assessing immunogenetic variation and the functional implications for immunity are fundamental toward understanding population-based disease associations.

Mammalian retrotransposons constitute 40% of the genome. During tissue regeneration, adult stem cells coordinately repress retrotransposons and activate lineage genes, but how this coordination is controlled is poorly understood. Here, we observed that dynamic expression of histone methyltransferase SETDB1 (a retrotransposon repressor) closely mirrors stem cell activities in murine skin. SETDB1 ablation leads to the reactivation of endogenous retroviruses (ERVs, a type of retrotransposon) and the assembly of viral-like particles, resulting in hair loss and stem cell exhaustion that is reversible by antiviral drugs. Mechanistically, at least two molecularly and spatially distinct pathways are responsible: antiviral defense mediated by hair follicle stem cells and progenitors and antiviral-independent response due to replication stress in transient amplifying cells. ERV reactivation is promoted by DNA demethylase ten-eleven translocation (TET)-mediated hydroxymethylation and recapitulated by ablating cell fate transcription factors. Together, we demonstrated ERV silencing is coupled with stem cell activity and essential for adult hair regeneration.

An antibody-based HIV-1 vaccine will require the induction of potent cross-reactive HIV-1-neutralizing responses. To demonstrate feasibility toward this goal, we combined vaccination targeting the fusion-peptide site of vulnerability with infection by simian-human immunodeficiency virus (SHIV). In four macaques with vaccine-induced neutralizing responses, SHIV infection boosted plasma neutralization to 45%-77% breadth (geometric mean 50% inhibitory dilution [ID50] ∼100) on a 208-strain panel. Molecular dissection of these responses by antibody isolation and cryo-electron microscopy (cryo-EM) structure determination revealed 15 of 16 antibody lineages with cross-clade neutralization to be directed toward the fusion-peptide site of vulnerability. In each macaque, isolated antibodies from memory B cells recapitulated the plasma-neutralizing response, with fusion-peptide-binding antibodies reaching breadths of 40%-60% (50% inhibitory concentration [IC50] < 50 μg/mL) and total lineage-concentrations estimates of 50-200 μg/mL. Longitudinal mapping indicated that these responses arose prior to SHIV infection. Collectively, these results provide in vivo molecular examples for one to a few B cell lineages affording potent, broadly neutralizing plasma responses.

Type III CRISPR systems provide immunity against genetic invaders through the production of cyclic oligo-adenylate (cAn) molecules that activate effector proteins that contain CRISPR-associated Rossman fold (CARF) domains. Here, we characterized the function and structure of an effector in which the CARF domain is fused to an adenosine deaminase domain, CRISPR-associated adenosine deaminase 1 (Cad1). We show that upon binding of cA4 or cA6 to its CARF domain, Cad1 converts ATP to ITP, both in vivo and in vitro. Cryoelectron microscopy (cryo-EM) structural studies on full-length Cad1 reveal an hexameric assembly composed of a trimer of dimers, with bound ATP at inter-domain sites required for activity and ATP/ITP within deaminase active sites. Upon synthesis of cAn during phage infection, Cad1 activation leads to a growth arrest of the host that prevents viral propagation. Our findings reveal that CRISPR-Cas systems employ a wide range of molecular mechanisms beyond nucleic acid degradation to provide adaptive immunity in prokaryotes.

Epithelial tumors are characterized by abundant inter- and intra-tumor heterogeneity, which complicates diagnostics and treatment. The contribution of cancer-stroma interactions to this heterogeneity is poorly understood. Here, we report a paradigm to quantify phenotypic diversity in head and neck squamous cell carcinoma (HNSCC) with single-cell resolution. By combining cell-state markers with morphological features, we identify phenotypic signatures that correlate with clinical features, including metastasis and recurrence. Integration of tumor and stromal signatures reveals that partial epithelial-mesenchymal transition (pEMT) renders disease outcome highly sensitive to stromal composition, generating a strong prognostic and predictive signature. Spatial transcriptomics and subsequent analyses of cancer spheroid dynamics identify the cancer-associated fibroblast-pEMT axis as a nexus for intercompartmental signaling that reprograms pEMT cells into an invasive phenotype. Taken together, we establish a paradigm to identify clinically relevant tumor phenotypes and discover a cell-state-dependent interplay between stromal and epithelial compartments that drives cancer aggression.

Differential expression analysis of single-cell RNA sequencing (scRNA-seq) data is central for characterizing how experimental factors affect the distribution of gene expression. However, distinguishing between biological and technical sources of cell-cell variability and assessing the statistical significance of quantitative comparisons between cell groups remain challenging. We introduce Memento, a tool for robust and efficient differential analysis of mean expression, variability, and gene correlation from scRNA-seq data, scalable to millions of cells and thousands of samples. We applied Memento to 70,000 tracheal epithelial cells to identify interferon-responsive genes, 160,000 CRISPR-Cas9 perturbed T cells to reconstruct gene-regulatory networks, 1.2 million peripheral blood mononuclear cells (PBMCs) to map cell-type-specific quantitative trait loci (QTLs), and the 50-million-cell CELLxGENE Discover corpus to compare arbitrary cell groups. In all cases, Memento identified more significant and reproducible differences in mean expression compared with existing methods. It also identified differences in variability and gene correlation that suggest distinct transcriptional regulation mechanisms imparted by perturbations.

Hippocampal CA3 is central to memory formation and retrieval. Although various network mechanisms have been proposed, direct evidence is lacking. Using intracellular Vm recordings and optogenetic manipulations in behaving mice, we found that CA3 place-field activity is produced by a symmetric form of behavioral timescale synaptic plasticity (BTSP) at recurrent synapses among CA3 pyramidal neurons but not at synapses from the dentate gyrus (DG). Additional manipulations revealed that excitatory input from the entorhinal cortex (EC) but not the DG was required to update place cell activity based on the animal's movement. These data were captured by a computational model that used BTSP and an external updating input to produce attractor dynamics under online learning conditions. Theoretical analyses further highlight the superior memory storage capacity of such networks, especially when dealing with correlated input patterns. This evidence elucidates the cellular and circuit mechanisms of learning and memory formation in the hippocampus.

Elucidating organismal developmental processes requires a comprehensive understanding of cellular lineages in the spatial, temporal, and molecular domains. In this study, we introduce Zebrahub, a dynamic atlas of zebrafish embryonic development that integrates single-cell sequencing time course data with lineage reconstructions facilitated by light-sheet microscopy. This atlas offers high-resolution and in-depth molecular insights into zebrafish development, achieved through the sequencing of individual embryos across ten developmental stages, complemented by reconstructions of cellular trajectories. Zebrahub also incorporates an interactive tool to navigate the complex cellular flows and lineages derived from light-sheet microscopy data, enabling in silico fate-mapping experiments. To demonstrate the versatility of our multimodal resource, we utilize Zebrahub to provide fresh insights into the pluripotency of neuro-mesodermal progenitors (NMPs) and the origins of a joint kidney-hemangioblast progenitor population.

Recent reports raise concerns on the changing epidemiology of mpox in the Democratic Republic of the Congo (DRC). High-quality genomes were generated for 337 patients from 14/26 provinces to document whether the increase in number of cases is due to zoonotic spillover events or viral evolution, with enrichment of APOBEC3 mutations linked to human adaptation. Our study highlights two patterns of transmission contributing to the source of human cases. All new sequences from the eastern South Kivu province (n = 17; 4.8%) corresponded to the recently described clade Ib, associated with sexual contact and sustained human-to-human transmission. By contrast, all other genomes are clade Ia, which exhibits high genetic diversity with low numbers of APOBEC3 mutations compared with clade Ib, suggesting multiple zoonotic introductions. The presence of multiple clade I variants in urban areas highlights the need for coordinated international response efforts and more studies on the transmission and the reservoir of mpox.

Mechanistic studies of life's lower metabolic limits have been limited due to a paucity of tractable experimental systems. Here, we show that redox-cycling of phenazine-1-carboxamide (PCN) by Pseudomonas aeruginosa supports cellular maintenance in the absence of growth with a low mass-specific metabolic rate of 8.7 × 10-4 W (g C)-1 at 25°C. Leveraging a high-throughput electrochemical culturing device, we find that non-growing cells cycling PCN tolerate conventional antibiotics but are susceptible to those that target membrane components. Under these conditions, cells conserve energy via a noncanonical, facilitated fermentation that is dependent on acetate kinase and NADH dehydrogenases. Across PCN concentrations that limit cell survival, the cell-specific metabolic rate is constant, indicating the cells are operating near their bioenergetic limit. This quantitative platform opens the door to further mechanistic investigations of maintenance, a physiological state that underpins microbial survival in nature and disease.

Pathogens constantly evolve and can develop mutations that evade host immunity and treatment. Addressing these escape mechanisms requires targeting evolutionarily conserved vulnerabilities, as mutations in these regions often impose fitness costs. We introduce adaptive multi-epitope targeting with enhanced avidity (AMETA), a modular and multivalent nanobody platform that conjugates potent bispecific nanobodies to a human immunoglobulin M (IgM) scaffold. AMETA can display 20+ nanobodies, enabling superior avidity binding to multiple conserved and neutralizing epitopes. By leveraging multi-epitope SARS-CoV-2 nanobodies and structure-guided design, AMETA constructs exponentially enhance antiviral potency, surpassing monomeric nanobodies by over a million-fold. These constructs demonstrate ultrapotent, broad, and durable efficacy against pathogenic sarbecoviruses, including Omicron sublineages, with robust preclinical results. Structural analysis through cryoelectron microscopy and modeling has uncovered multiple antiviral mechanisms within a single construct. At picomolar to nanomolar concentrations, AMETA efficiently induces inter-spike and inter-virus cross-linking, promoting spike post-fusion and striking viral disarmament. AMETA's modularity enables rapid, cost-effective production and adaptation to evolving pathogens.

Inflammatory cytokines are pivotal to immune responses. Upon cytokine exposure, cells enter an "alert state" that enhances their visibility to the immune system. Here, we identified an alert-state subpopulation of ribosomes defined by the presence of the P-stalk. We show that P-stalk ribosomes (PSRs) are formed in response to cytokines linked to tumor immunity, and this is at least partially mediated by P-stalk phosphorylation. PSRs are involved in the preferential translation of mRNAs vital for the cytokine response via the more efficient translation of transmembrane domains of receptor molecules involved in cytokine-mediated processes. Importantly, loss of the PSR inhibits CD8+ T cell recognition and killing, and inhibitory cytokines like transforming growth factor β (TGF-β) hinder PSR formation, suggesting that the PSR is a central regulatory hub upon which multiple signals converge. Thus, the PSR is an essential mediator of the cellular rewiring that occurs following cytokine exposure via the translational regulation of this process.

Endo-β-N-acetylglucosaminidases (ENGases) that specifically hydrolyze the Asn297-linked glycan on immunoglobulin G (IgG) antibodies, the major molecular determinant of fragment crystallizable (Fc) γ receptor (FcγR) binding, are exceedingly rare. All previously characterized IgG-specific ENGases are multi-domain proteins secreted as an immune evasion strategy by Streptococcus pyogenes strains. Here, using in silico analysis and mass spectrometry techniques, we identified a family of single-domain ENGases secreted by pathogenic corynebacterial species that exhibit strict specificity for IgG antibodies. By X-ray crystallographic and surface plasmon resonance analyses, we found that the most catalytically efficient IgG-specific ENGase family member recognizes both protein and glycan components of IgG. Employing in vivo models, we demonstrated the remarkable efficacy of this IgG-specific ENGase in mitigating numerous pathologies that rely on FcγR-mediated effector functions, including T and B lymphocyte depletion, autoimmune hemolytic anemia, and antibody-dependent enhancement of dengue disease, revealing its potential for treating and/or preventing a wide range of IgG-mediated diseases in humans.