Cell

Linking cellular phenotype to genotype and gene expression is a central goal in biomedical research. Live-cell imaging captures dynamic cellular phenotypes, whereas single-cell sequencing measures genome-wide molecular states, but pairing these readouts for the same cell remains challenging. Here we present Waldo, a genetically encoded single-cell barcoding strategy that can be read out by both live-cell imaging and single-cell sequencing to connect single cells across both modalities. Waldo barcodes are generated by combinatorial expression of synthetic cell-surface receptors, each detected by a cognate peptide probe coupled either to a fluorescent dye or a DNA barcode. Iterative fluorescent probe staining achieves large barcode diversity, while preserving fluorescence channels for phenotypic reporters. As a proof-of-principle, we use Waldo to link cell division phenotypes after chromosome missegregation to the karyotype and transcriptome of the same cells. Waldo provides a general strategy for connecting imaging phenotypes to genome-wide sequencing readouts in single cells.

Bacteria commit to costly behaviors before they can benefit from them. In Escherichia coli, flagellar synthesis is expensive, yet the genes are expressed in stochastic pulses and only a small minority of pulses result in motility. Despite this noise, we find that lineage-related cells begin swimming at similar times, in groups of two to eight cells. Rare, high-amplitude pulses commit a single cell within one generation, but more commonly, pulses rise slowly and fail to trigger swimming before division. Rather than decaying, their output accumulates as cells inherit hook-basal bodies (HBBs) and continue to build new ones. This structural inheritance acts as a physical integrator, summing ongoing transcriptional activity across generations, so that an incomplete flagellar cascade is carried forward at division rather than lost. HBB accumulation sets when cells swim, and the anti-sigma factor FlgM sets how sharply. By holding late flagellar genes off until several HBBs have accumulated, FlgM makes motility depend steeply on early flagellar gene expression, so that relatives inheriting similar sets of HBBs become motile together. Without FlgM, this sharp dependence is largely lost, and activations of the full cascade do not reliably produce swimming. Together, our results show how the accumulation of a molecular machine integrates fluctuations, and how a checkpoint on its assembly sharpens them into a decision, setting both the speed and heritability of a behavioral trait.

T cells persist in the central nervous system (CNS) and can drive both protection and neurological disease. How these cells are organized in humans and what they recognize is largely unknown. Here, we profiled CD8 T cells across anatomically distinct CNS regions, obtained through on-site autopsies and temporal lobe resection surgeries, using single-cell RNA sequencing, paired T cell receptor sequencing, and DNA-barcoded tetramers. Resident memory T cells (TRM) specific for Epstein-Barr virus, cytomegalovirus, influenza A, and SARS-CoV-2 were identified across CNS compartments. Anatomical location was the strongest correlate of TRM cell state, with leptomeningeal cells adopting a cytokine-poised TRM program, whereas brain TRM cells were transcriptionally restrained. Cells of the same clonotype spanned tissues yet adopted local transcriptional states. Viral specificity added another layer of TRM heterogeneity with GZMK/GZMA-expressing EBV-specific populations and interferon-stimulated gene signatures in SARS-CoV-2 and Influenza A-specific cells. The human CNS thus harbors regionally distinct CD8+ TRM shaped by common viral exposures.

Viral development links gene expression and genome replication to particle formation within a cell. Here we construct virtual viruses in a JCVI-syn3A-based spatial minimal cell. For mycoplasma virus P1, molecular events connect expression and replication to coarse-grained packaging, followed by a spatial reconstruction of membrane opening and particle release. Individual finite-size particles coexist, move across the boundary and undergo partial crossings and returns before complete exit. In contrast, Acholeplasma phage MV-L1 (L1) expresses its genes and accumulates genomes as the host continues to grow and develops a constricted shape. The coexistence of viral activity and continued host growth is qualitatively consistent with the non-lytic biology of group 1 acholeplasmaviruses. These reconstructions extend the virtual cell to the viruses that develop within it, linking molecular activity, particle behavior and host growth.

Many epigenetic regulators play essential roles in animal development, but precisely how they function across different cell types without clear sequence preferences remains a foundational challenge. Here, we employed single-cell RNA sequencing to survey the mutant phenotypes of 30 essential regulatory pathways across over 450 gastrulation stage mouse embryo replicates, including essential factors associated with repressive chromatin states, RNA surveillance/processing, and other genomic roles. We then use this integrated manifold to partition mutant responses into a developmentally-resolved interaction network, revealing a hierarchy of heterochromatin-associated interactions that act orthogonally to Polycomb Repressive Complex-mediated control of embryo patterning. Compared to the embryo proper, we find that the developing placenta and yolk sac can accommodate broad swings in repetitive element expression, a resilience in keeping with the relative evolutionary recency and phenotypic plasticity of these tissues. Dense temporal sampling of Human Silencing Hub (HUSH) mutant embryogenesis further supports the notion that the gastrulating embryo is particularly constrained and depends on multiple layers of insulation, including the need to suppress evolutionarily young genetic elements to stably license highly conserved, phylotypic programs. In combination, our work highlights the power of single cell-resolved developmental genetics to contextualize the roles of epigenetic regulators, which restrict genomic action to support robust cellular identities and embryological forms.

Viral-induced DNA damage triggers both eukaryotic and prokaryotic innate immunity. Here, we report that self-DNA ends generated routinely during normal termination of bacterial replication suffice to trigger the P2-encoded OLD antiphage system. Normally, these physiological triggers are cleared by the dual helicase-nuclease activities of the conserved RecBCD enzyme, which acts as an immune safeguard analogous to mammalian TREX1 exonuclease and plant DBR1 debranching enzyme. However, when this clearance fails, the transient self-DNA ends persist and are recognized by ATP-bound OLD dimers via a positively charged channel across their ATPase subunits. ATP hydrolysis then activates OLD to cleave tRNAs, causing cytotoxicity. Single-cell microscopy reveals that dozens of OLD molecules associate with kilobases of adjacent DNA, amplifying the sparse physiological DNA end signal. Our findings reveal how an endogenous byproduct poses a constant autoimmune threat in bacteria, highlighting a universal molecular conflict across kingdoms and the convergent solutions evolved to resolve it.

Mycobacterium tuberculosis (Mtb) secretes antigenic proteins that drive protective T cell responses, yet how Mtb intracellular localization affects the trafficking and spatiotemporal distribution of mycobacterial antigens in macrophages remains elusive. Here, we developed a split fluorescent protein reporter to monitor the dynamics of the TB vaccine lead antigen Ag85B in infected human macrophages. We discovered that Mtb triggers the formation and remodeling of two distinct types of Mtb Antigen Storage Compartment (MASC) that retains Ag85 within macrophage membranes. Using subcellular spatial proteomics, we profiled the composition of the MASC, identifying both the Mtb and macrophage proteins that constitute these compartments. Unexpectedly, MASCs are defined by metabolic compartmentalization and rewiring to lipid metabolism in both the macrophage and Mtb. Thus, Mtb builds via the ESX-1 system a metabolically distinct antigen storage compartment in immune cells, with potential implications for antigen selection and delivery in TB vaccine development.

The emergence of reproductive barriers between closely related populations is a fundamental step in speciation, yet the underlying genetic mechanisms remain poorly understood in mammals. Here we resolve the DDK syndrome, a long-standing model of hybrid incompatibility in mice, by showing that two independent retrotransposon insertions rewire immune Schlafen endoribonuclease genes expression, causing hybrid embryo death. A DDK-private MERVL element drives ectopic oocyte expression of Slfn1, while an intragenic ETn/MusD insertion in incompatible strains is associated with zygotic Slfn15 expression. SLFN1-SLFN15 heterodimerization constitutively activates tRNA ribonuclease activity, leading to global translation inhibition, and integrated stress response activation. Abolishing SLFN1 catalytic activity or SLFN15 expression rescues syndromic embryo development. Our findings uncover a previously unrecognized speciation mechanism in which transposable elements repurpose immune effectors as drivers of postzygotic reproductive isolation.

Cadmium (Cd) is an emerging environmental risk factor for cognitive decline and Alzheimer's disease, and few interventions limit its effects on the brain. Akkermansia muciniphila, a human gut commensal, is neuroprotective through mechanisms that remain unclear. Building on our finding that A. muciniphila prevented Cd-induced memory deficits without reducing brain Cd burden (Wang et al., 2026a), we tested whether it reprograms Cd-disrupted hippocampal RNA regulation. Male C57BL/6J mice received A. muciniphila (1x10^9 CFU, oral gavage, 5 days/week) before and throughout 9 weeks of 3 mg/L CdCl2 exposure; hippocampi were analyzed by RNA-seq, m6A epitranscriptomic profiling, and rMATS-based alternative-splicing analysis. Cd disrupted hippocampal neuropeptide signaling, neuronal excitability, and synaptic plasticity. Co-exposure with A. muciniphila reversed signatures enriched for Nrf2-centered cytoprotective signaling and opposed m6A and splicing changes across more than a thousand genes, largely without changes in total mRNA abundance. Restored events converged on machinery governing local translation, calcium-dependent plasticity, and synaptic transmission. Projected onto predicted protein structures, most introduced a premature termination codon predicted to trigger nonsense-mediated decay or edited short loop and linker segments rather than rebuilding folded domains, regulation largely invisible to gene-level expression analysis. Among these, paired events amplifying semaphorin-plexin signaling, an axis linked to oligodendrocyte injury and memory consolidation, were reversed. These findings identify restoration of hippocampal transcript fate and splice-isoform selection as a mechanism potentially underlying A. muciniphila-mediated protection against Cd neurotoxicity, and establish post-transcriptional RNA regulation as a gut-brain target for mitigating environmentally driven neurotoxicity.

Ayahuasca, used religiously for centuries by Amazonian Indigenous peoples and increasingly studied for its therapeutic potential, owes its psychoactivity to two plants: Psychotria viridis supplies N,N-dimethyltryptamine (DMT), whereas Banisteriopsis caapi supplies beta-carbolines that render DMT orally active. Here we show that the two plants have distinct genome histories: P. viridis is hexaploid (2n = 6x = 66), consistent with autopolyploidy, and B. caapi is diploid (2n = 2x = 20). CENH3-bound regions of both species are nonetheless enriched in CRM-type Ty3/Gypsy retrotransposons, with additional LINE elements and CRM-derived satellites in B. caapi. In P. viridis, a single multifunctional methyltransferase catalyzes both N-methylations of tryptamine to DMT. In B. caapi, beta-carboline biosynthesis combines dedicated enzymes with an acetaldehyde-dependent Pictet-Spengler cyclization that can proceed non-enzymatically, followed by enantioselective oxidation to harmaline. These resources link the ethnobotanical history of ayahuasca to its molecular basis, providing genomes of both source plants, the genetic structure of traditionally recognized cultivated varieties, and the enzymes behind the alkaloids long combined in this decoction.

Bacteria and archaea encode hundreds of antiviral defense systems, some with homologs found in eukaryotic innate immunity. GHL ATPases like plant MORCs and prokaryotic Tiamat defend against viruses and phage, though their architecture and biochemical activities have remained unclear. Searching thousands of proteomes, we find Tiamat is rare in prokaryotes yet enriched across eukaryotes, including green plants, aquatic chordates, sponges, placozoans, and water molds. We resolve the central DUF3684 and flanking region into three predicted folds (TPR-like repeats, an uncharacterized conserved fold, and an inactive metalloprotease-like fold) that occupy the substrate-binding position of other GHL proteins. We show that purified Bacillus cereus Tiamat hydrolyzes ATP, inefficiently nicks duplex DNA, and cleaves single-stranded DNA site-specifically. We identify Tiamat-like proteins whose C-terminal nuclease is often swapped for helicases, peptidases, or other effectors. Additionally, we show Tiamat genes neighbor helicases, DNA methyltransferases, and nucleases, suggesting some Tiamats degrade foreign nucleic acid within a multi-component complex. Eukaryotic Tiamat homologs like the close Arabidopsis homolog NO VEIN and the distant human homolog SACSIN have acquired critical non-defense roles, leaving open whether these homologs remain antiviral or have been exapted.

Aerial roots of certain maize (Zea mays) and sorghum (Sorghum bicolor) genotypes secrete a carbohydrate-rich mucilage that sustains nitrogen-fixing bacteria, and the underground roots of most plants secrete mucilage at their tips. In both settings, border cells released from the root cap produce this mucilage, yet the genes controlling their formation remain largely unknown. We sequenced the RNA of about 23,000 nuclei from mucilage-secreting roots of maize, sorghum, and Medicago truncatula, sampled before and after secretion. Each species yielded 10 to 15 clusters, which we annotated with published markers and with new markers that we validated by in situ hybridization in the two cereals and by promoter-reporter fusions in M. truncatula. We then inferred modular and cell type-specific gene regulatory networks. Transcriptional programs diverged across the three species except in the quiescent center and the meristem, and in the clusters containing border cells, where a shared set of regulators emerged. Silencing three of five candidate regulators reduced border cell number, while silencing a partially overlapping set of three regulators reduced root diameter in M. truncatula. Silencing the C2H2 zinc finger protein MtrunA17_Chr1g0148421 also reduced border cell viability. A maize insertion mutant of bzip89 released fewer border cells than the wild type. These networks offer entry points for tuning mucilage production in cereal and legume crops.

Metagenomic databases hold vast quantities of unexplored genetic material that can seed tools for biotechnology and medicine, but discovery is limited by our ability to examine sequence data at scale. Here, we show that large language models (LLMs) in an autonomous harness can alleviate this bottleneck and surface novel molecular systems. We deployed Claude Code instances to survey reverse transcriptase (RT) loci across 1.9 billion protein clusters. Among the top candidates were array-associated RTs (ART), a novel family of jumbo-phage RTs coupled with an array of ~200-nucleotide units and a dedicated partner gene. Session transcripts show that Claude identified the ART family by directly examining its DNA sequences and recognizing the atypical repeats, a behavior attributable to specific Mythos 5 internal signals that respond to repeated DNA. ART arrays are highly expressed and appear as discrete units during Staphylococcus phage infection, suggesting an RT system directed by a repertoire of distinct RNAs. These findings indicate that LLMs can autonomously detect anomalies and drive analyses to initiate biological discoveries.

The human spermatozoon is among the most specialized cells in the body, yet where its proteins reside, where they come from and how they vary between and within men remain largely unresolved. Here we present the deepest proteomic atlas of human spermatozoa to date, identifying 10,828 proteins across six donors. Abundances correlate well between individuals and are anchored by a conserved core of mitochondrial energy-metabolism proteins, while a smaller fraction varies between donations from the same man. By mechanically separating heads from tails, we assigned thousands of previously unlocalized proteins to a compartment, and by comparing seminal fluid from vasectomized and non-vasectomized men we distinguished sperm- and epididymis-derived proteins from accessory-gland secretions. Using a single-cell analysis workflow we further show that the proteomes of individual spermatozoa can be measured, recovering up to ~1,000 protein groups per cell. In a cohort with total fertilization failure, we find recurrent mitochondrial rather than sperm-egg fusion dysregulation. Together or data provides novel insights into the dynamics of the sperm proteome and male infertility.

Every embryo inherits a maternal proteome that it must remodel with zygotic proteins to build its many cell types. The fate of the maternal proteome remains contested because indirect measurements cannot resolve it. Here, we combine 18O-water labeling with multiplexed proteomics to quantify protein turnover proteome-wide in frog and fly embryos. Through hatching, the frog preserves the bulk of its maternal proteome, confining rapid degradation to a small regulatory module. The fly cannot meet its synthesis demand from yolk alone and instead degrades nearly all maternal proteins, including housekeeping proteins long assumed stable, recycling them into new protein. Yet the turnover hierarchy is conserved, with disordered and regulatory proteins degrading fastest, while the fly rescales the whole proteome ~eightfold faster. These results recast the developmental proteome as both informational inheritance and metabolic reserve, establish 18O-water labeling as a turnover method for non-feeding organisms, and provide a resource of embryonic half-lives.

Human leukocyte antigen B*27 (HLA-B*27) is a major risk factor for autoimmune diseases such as axial spondyloarthritis (axSpA), acute anterior uveitis (AAU), and psoriatic arthritis (PsA). HLA-B*27 presents bacterial and self-antigens to disease-associated CD8+ T cells, however, how this leads to a pathogenic IL-17-mediated disease remains unclear. We performed mutational experiments in an established disease-associated TCR and demonstrated that the pathogenic motif encompasses a wider range of TCR sequences than previously appreciated. We next leveraged computational pairing of alpha/beta TCR sequencing (TCR-seq) to demonstrate that the broadened pathogenic motif distinguishes axSpA and AAU patients from healthy controls. Paired single-cell RNA sequencing and single-cell TCR sequencing datasets demonstrated that the pathogenic cells have a distinct transcriptional signature. We find a coupling of pathogenic TCRs to this transcriptional signature in public datasets from axSpA and PsA joint fluid and AAU ocular fluid. Finally, we detect the pathogenic CD8+ T cells in the gut and demonstrate that they express a Type 17 program. Our findings suggest that HLA-B*27+ axSpA, AAU, and PsA are initiated by HLA-B*27-restricted pathogenic CD8+ T cells that undergo Type 17 differentiation in response to intestinal antigens.

How a flexible mammalian embryo gives rise to a reproducible body plan remains unclear, partly because cell lineages cannot be followed continuously in vivo. We combined DNA Typewriter lineage recording with tetraploid complementation to reconstruct high-resolution single-cell genealogies across mouse development and built FateVec, an analysis framework that infers fate-bias dynamics and lineage-associated transcriptional programs from reconstructed phylogenies. Loss of developmental potential proved progressive and asynchronous, with multiple lineages acquiring detectable fate bias before morphological diversification at gastrulation. Within the neural crest, most cells remained multipotent over an extended window, while independent founder clades differed markedly in restriction timing. Extending lineage recording to a stem-cell-derived embryo model revealed that cells with shared transcriptional identities can arise from distinct lineage relationships. We further found embryonic-lineage contributions to extraembryonic-like states, exposing an unexplored lineage plasticity. This lineage-resolved view of embryogenesis reveals that cell identity alone predicts neither developmental ancestry nor the timing of fate restriction.

The oral cavity harbors one of the body's most diverse microbial communities, which continuously seeds the bloodstream and gastrointestinal tract, particularly in periodontitis, a disease in which both barrier permeability and microbial burden increase. Despite this constant systemic exposure, overt bacteremia from oral organisms is rare, suggesting an active surveillance system that contains this reservoir. Here, we apply IgG-sequencing to comprehensively map mucosal and systemic antibody recognition of the oral microbiome across paired oral fluid and serum from healthy individuals and patients with periodontitis. We find that IgG-binding is only weakly predicted by microbial abundance, suggesting taxon-specific and largely conserved IgG-binding of commensals, many of them streptococci which dominate antibody recognition irrespective of disease state. Periodontitis broadens the repertoire of IgG-bound taxa to include particular disease-associated pathobionts while preserving robust targeting of health-associated symbionts. Comparison of salivary and serum antibody profiles further uncovers a subset of commensal streptococci, including species implicated in infective endocarditis, that acquire disproportionate systemic IgG binding in periodontitis. Together, these findings establish a taxon-level atlas of humoral immunosurveillance at a major mucosal-microbial interface and reveal how periodontal disease reshapes systemic antibody recognition of both pathobionts and commensals.

As climate change accelerates, the microbial communities sustaining organisms and ecosystems worldwide face an uncertain future, with some proving resilient to stress while others collapse, yet no generalizable framework exists to distinguish the two beforehand. Using amplicon sequencing data from ~29,000 microbiomes, we constructed over 440 matched network pairs spanning three of the most pressing stressors: drought, warming, and salinity. We show that keystone microbes with high connectivity, functional redundancy, and metabolic flexibility, are disproportionately likely to survive stress, and that baseline network and functional properties predict community-level structural and functional stability even as taxonomic composition shifts stochastically. These relationships held across the diverse hosts, ecosystems, and regions we tested. Because these properties can be derived from standard, low-cost sequencing collected before disturbance occurs, our findings turn baseline microbiome data into an early-warning system for identifying which communities are most vulnerable to climate stress.

Cell-autonomous immunity against gram-negative bacterial pathogens relies on Guanylate-binding proteins (GBPs). GBPs are interferon-inducible guanosine triphosphate hydrolases that coat cytosolic bacteria and promote non-canonical inflammasome activation. GBP coat assembly is initiated by GBP1, which binds bacterial lipopolysaccharide and recruits further paralogs in a hierarchical and nucleotide-dependent manner. Dissecting the contributions of individual GBPs has remained challenging because the seven human paralogs are structurally similar and undergo nucleotide-driven conformational changes that are difficult to resolve in cellular and biochemical experiments. Here, we developed a nanobody toolbox for the state-selective probing of human GBP coat assembly. We identify eight nanobodies with distinct paralog, domain, and conformational specificities and define their binding profiles and epitopes using complementary biophysical assays and cryogenic electron microscopy. We further demonstrate the utility of selected nanobodies by demonstrating their functionality in membrane reconstitution assays and in infected cells. A dimer-disrupting nanobody blocks progression of GBP-coat assembly from soluble GBP1 assemblies to LPS-associated coats, suggesting a stage-specific structural transition during coat formation. In contrast, a dimer selective nanobody binds assembled coats without measurably perturbing their organisation and is robustly recruited to GBP1-positive bacteria in infected cells, indicating that this dimer interface remains intact and accessible in coat-associated GBP1. Together, these nanobodies establish a set of state-selective reagents to distinguish, perturb, and probe GBP assemblies in structural and cellular studies.