Nature

Studying monozygotic twins who present with identical cancers informs on the developmental origins of childhood tumours. Here, we performed whole genome sequencing on multiple tumour, normal, and placental samples to reconstruct the phylogeny of a soft tissue cancer that spread in utero between monozygotic twins. This generalisable and scalable approach allows us to dissect the earliest stages of twinning, revealing unexpected asymmetrical contributions of embryonic lineages to the placenta and each twin.

Large language models embedded in autonomous agents process trusted instructions and untrusted data in one context window, leaving them open to direct and indirect prompt injection. In healthcare this is not hypothetical: a 2025 JAMA Network Open study found commercial medical LLMs followed injected instructions in 94.4% of simulated patient encounters, including life threatening recommendations . Yet the clinically decisive problem we quantify here is different. Most real clinical threats protected health information PHI exfiltration, cross patient access, bulk export, out of scope advice are fluent, legitimate looking requests that carry no attack signal, so even a state of the art injection detector passes them. Existing runtime guardrails trade safety against latency: model based auditors are accurate but add hundreds of milliseconds of Python inference, while lexical filters are fast but blind to obfuscated or semantically disguised payloads. We present QFIRE, an inline, provider agnostic prompt firewall implemented as a single self contained Rust toolchain proxy, CLI, and benchmark harness. QFIRE combines three mechanisms: (i) positive security scope constraints, which restrict a model call to a declared natural language purpose and block out of scope drift even when no overt attack token is present; (ii) an asynchronous detector graph that runs N rules and their detector nodes concurrently, cheapest checks first; and (iii) a de obfuscation pass that decodes Base64 hex ROT13, folds homoglyphs and leetspeak, and strips zero width characters before detection. QFIRE ships 106 versioned firewall rules and a dedicated HIPAA Safe Harbor 18 identifier PHI panel, and runs a local DeBERTa v3 injection classifier via embedded ONNX Runtime. On 1968 public prompt injection and jailbreak prompts QFIREs deterministic hybrid attains F1 0.86, statistically tied with Metas state of the art PromptGuard 2 0.86 and above protectai DeBERTa v3 0.83; lexical baselines lag 0.16 to 0.50. Our central result is on QFIRE HealthBench, a new 2000 prompt healthcare benchmark we build and release with real garak and Microsoft PyRIT payloads. There the same PromptGuard-2 recovers only 0.40 recall DeBERTa v3 0.57, because most clinical threats carry no injection signal; QFIREs combined scope plus PHI chain reaches 0.83 recall F1 0.87 at a calibrated 0.08 false positive rate. Generic injection detection, even state of the art, is therefore necessary but not sufficient for healthcare agents. A bare LLM judge also closes most of this static corpus gap F1 0.90; QFIREs contribution beyond static accuracy is auditable determinism, bounded latency, and adaptive robustness, where the bare judge falls to 34 to 59% recall section 5.5. End to end, placing QFIRE in front of a tool using agent over a mock EHR sandbox cuts the agents harmful action rate from 0.38 to 0.00 at a 0.13 benign utility cost. All code, rules, corpora snapshots, and scripts are released, and every table regenerates from a single make paper target against local models with no paid API keys.

The assembly of functional neural circuits relies on the generation of diverse neural types with precise molecular identity and connectivity. Unlocking general principles of neuronal specification and wiring across the nervous system requires a systematic and high-resolution characterisation of its diversity, recently enabled by advances in single-cell transcriptomics and connectomics. However, linking the molecular identity of neurons to circuit architecture remains a key challenge. Here, we present a high-resolution developmental transcriptional atlas for the Drosophila melanogaster nerve cord, the central hub for sensory-motor circuits. With an unprecedented 38x aggregate coverage relative to its reference connectome1,2, our atlas captures extensive molecular diversity and enables robust alignment to the adult connectome. We identified three developmental principles underlying neuronal diversity in the nerve cord. First, timing of neurogenesis shapes diversification of molecular identity: embryonic-born neurons diverge faster than larval-born neurons, as also observed in the adult connectome. Second, 17 transcription factors common to neurons from all lineages provide a global molecular identity code for birth order. Lastly, by mapping sex-specific transcriptional profiles to the connectome, we uncovered female-specific apoptosis and transcriptional divergence as key global drivers of sex-specification. By revealing key organisational axes of molecular identity, this atlas opens new avenues to dissect the molecular mechanisms underpinning the development and evolution of neural circuits.

Just as genomes revolutionized molecular genetics, connectomes (maps of neurons and synapses) are transforming neuroscience. To date, the only species with complete connectomes are worms1-3 and sea squirts4 (103-104 synapses). By contrast, the fruit fly is more complex (108 synaptic connections), with a brain that supports learning and spatial memory5,6 and an intricate ventral nerve cord analogous to the vertebrate spinal cord7-11. Here we report the first densely reconstructed adult fly connectome that unites the brain and ventral nerve cord, and we leverage this resource to investigate principles of neural control. We show that effector neurons (motor neurons, endocrine cells and efferent neurons targeting the viscera) are primarily influenced by sensory neurons in the same body part, forming local feedback loops. These local loops are linked by long-range circuits involving ascending and descending neurons organized into behavior-centric modules. Single ascending and descending neurons are often positioned to influence the voluntary movements of multiple body parts, together with the endocrine cells or visceral organs that support those movements. Brain regions involved in learning and navigation supervise these circuits. These results reveal an architecture that is distributed, parallelized and embodied, reminiscent of distributed control architectures in engineered systems12,13.

Task fMRI1 and electrophysiology2 have revealed distributed, linked cortical patches with shared category preferences (e.g., faces, objects, places)1,3-5, smaller than cytoarchitectonic areas. Resting-state functional connectivity (RSFC) similarly showed that somato-cognitive action network (SCAN) nodes interleave with effectors (foot, hand, mouth), subdividing the precentral gyrus6. Here, using multiple precision functional mapping (PFM) modalities (RSFC, task, lags), we discovered that most of association cortex is organized like face processing and SCAN, with small, discrete patches interconnected into chains. Such patch-chains densely tile prefrontal cortex but are largely absent from primary cortex. Cortico-striatal connectivity is organized such that patches of the same chain connect to the same striatal location. Within chains, infra-slow fMRI signals are ordered in time. RSFC-defined chains align with task fMRI localizers (e.g., visual, motor, pain). Chains are absent at birth and emerge in the first year of life, suggesting their formation is at least partially experience-driven. Cytoarchitectonic areas are subdivided by patches, and patches in the same chain are distributed across different cytoarchitectures. Chains represent parallel ordered processing streams that are separated by information domain and behavioral goals, not cytoarchitectonics. Functional subdivision of architectonics into smaller patches, interlinked to form cross-architecture chains, enable greater parallelization and flexible specialization of processing.

Animal and bacterial cells defend against viral infection by rapidly activating antiviral restriction factors. In human cells, antiviral immunity is initiated by interferon signaling that results in expression of hundreds of interferon-stimulated genes (ISGs)1,2,3,4,5,6. Complex regulatory networks and co-evolution of viral evasion strategies complicate analysis of immune proteins under native conditions and the function of most individual ISGs remains unknown7,8,9,10,11,12. Here we discover that heterologous expression of ISGs in bacteria is sufficient to protect against infection by diverse bacteriophages demonstrating properties of antiviral restriction preserved across billions of years of viral evolution. A screen of 306 human ISGs against 11 E. coli phages reveals that ISGs can restrict phage replication with potency equal to endogenous bacterial defense systems. We select for phage mutants that escape ISG restriction and identify the phage DNA primase-helicase complex as a target of human SPSB1. A 2.0 [A] crystal structure of the SPSB1-primase complex uncovers a recognition mechanism of foreign DxNxN protein motifs found in many animal and bacterial viral replication proteins. We show that SPSB1 in human cells recognizes and induces degradation of protein targets containing this motif from norovirus and poxvirus pathogens. Our results establish a cross-kingdom approach to studying immune function and reveal that human ISGs target features of viral replication common across kingdoms of life.

Oncogene amplification on extrachromosomal DNA (ecDNA) is a common driver of tumor progression and is associated with acquired drug resistance and poor patient survival. While whole genome sequencing (WGS) studies have revealed the landscape of genes amplified on ecDNA in tumors, it remains challenging to study the subclonal heterogeneity and functional (e.g., transcriptomic) consequences of ecDNA on tumors. To address this, we introduce scAmp: a probabilistic algorithm for detecting and analyzing ecDNA from single-cell datasets. We demonstrate scAmps improved accuracy over WGS approaches on well-characterized cell-lines and its applicability to clinical histopathology. We further showcase scAmp by analyzing 73 patient tumors profiled with single-cell ATAC-seq, where we analyze the subclonal evolution of ecDNA+ subclones and identify the effect of ecDNA amplifications on the chromatin accessibility landscape of cancer cells. Together, we anticipate that scAmp will broadly enable further studies - both retrospective and prospective - that dissect critical questions of how ecDNA affect cancer cells and the tumors in which they reside.

The function of the mammalian brain relies upon the specification and spatial positioning of diversely specialized cell types. Yet, the molecular identities of the cell types, and their positions within individual anatomical structures, remain incompletely known. To construct a comprehensive atlas of cell types in each brain structure, we paired high-throughput single-nucleus RNA-seq with Slide-seq-a recently developed spatial transcriptomics method with near-cellular resolution-across the entire mouse brain. Integration of these datasets revealed the cell type composition of each neuroanatomical structure. Cell type diversity was found to be remarkably high in the midbrain, hindbrain, and hypothalamus, with most clusters requiring a combination of at least three discrete gene expression markers to uniquely define them. Using these data, we developed a framework for genetically accessing each cell type, comprehensively characterized neuropeptide and neurotransmitter signaling, elucidated region-specific specializations in activity-regulated gene expression, and ascertained the heritability enrichment of neurological and psychiatric phenotypes. These data, available as an online resource (BrainCellData.org) should find diverse applications across neuroscience, including the construction of new genetic tools, and the prioritization of specific cell types and circuits in the study of brain diseases.

Since sensory information is always present in the environment, animals need to internally regulate their responsiveness to fit the context. During sleep, the threshold for sensory arousal is increased so that only stimuli of sufficient magnitude can cross it. The mechanisms that make arousability flexible are largely mysterious, but they must integrate sensory information with information about physiology. We discovered a gut-to-brain signaling pathway that uses information about ingested nutrients to control arousability from sleep, without affecting sleep duration. Protein ingestion causes endocrine cells in the Drosophila gut to increase production of CCHa1, a peptide that decreases sensory responsiveness. CCHa1 is received by a small group of brain dopaminergic neurons whose activity gates behavioral responsiveness to mechanical stimulation. These dopaminergic neurons innervate the mushroom body, a brain structure involved in determining sleep duration. This work describes how the gut tunes arousability according to nutrient availability, allowing deeper sleep when dietary proteins are abundant. It also suggests that behavioral flexibility is increased through independent tuning of sleep depth and duration.

Mammalian brain function relies on the precise synaptic architecture of diverse cell types, yet scalable methods for linking a neurons transcriptomic profile to its neuroanatomy remain limited. We present Synapse-seq, an in vivo strategy in which cell-identifying barcoded mRNAs are routed to subcellular compartments via targeting proteins and detected by single-cell and spatial genomics. Using AAV delivery for minimal perturbation of gene expression, we directed barcodes to presynaptic terminals (via synaptophysin) in four distinct circuits, or to postsynaptic sites (via nanobodies to endogenous PSD95) of hippocampal excitatory neurons. In the mouse primary visual cortex, presynaptic Synapse-seq recovered known long-range projections and discovered cortical layer subtypes with distinct thalamic innervation. In the anterior cortex, we elucidated simple topographic rules of corticostriatal innervation: intratelencephalic neurons followed a continuous depth-to-target gradient, while extratelencephalic neurons exhibited striatal collaterals that spatially correlated with medullary innervation. Finally, postsynaptic barcoding of excitatory neurons revealed cell type-specific variation in dendritic architectures across and within hippocampal subfields. These data establish Synapse-seq as a versatile, genomics-based approach for the integrated definition of molecular identity and synaptic organization across mammalian brains.

Many animals rely on vision to navigate through their environment. The pattern of changes in the visual scene induced by self-motion is the optic flow1, which is first estimated in local patches by directionally selective (DS) neurons2-4. But how should the arrays of DS neurons, each responsive to motion in a preferred direction at a specific retinal position, be organized to support robust decoding of optic flow by downstream circuits? Understanding this global organization is challenging because it requires mapping fine, local features of neurons across the animals field of view3. In Drosophila, the asymmetric dendrites of the T4 and T5 DS neurons establish their preferred direction, making it possible to predict DS responses from anatomy4,5. Here we report that the preferred directions of fly DS neurons vary at different retinal positions and show that this spatial variation is established by the anatomy of the compound eye. To estimate the preferred directions across the visual field, we reconstructed hundreds of T4 neurons in a full brain EM volume6 and discovered unexpectedly stereotypical dendritic arborizations that are independent of location. We then used whole-head CT scans to map the viewing directions of all compound eye facets and found a non-uniform sampling of visual space that explains the spatial variation in preferred directions. Our findings show that the organization of preferred directions in the fly is largely determined by the compound eye, exposing an intimate and unexpected connection between the peripheral structure of the eye, functional properties of neurons deep in the brain, and the control of body movements.

Kinship practices underpin all traditional societies, forming the basis for socially sanctioned reproductive unions, residence patterns and the inheritance of rights and property1,2. Although the relationship between biological relatedness and kinship is not always straightforward, ancient DNA studies are increasingly used to examine the extent to which biological relatedness underpinned social constructs of kinship in prehistoric societies3-6. Here, we report the analysis of genome-wide data for 534 individuals from the Arras Culture of Middle Iron Age northeast England (including 390 from Wetwang Slack, 100 from Pocklington, and 29 from Melton), finding evidence for communities with kinship systems structured along matrilineal lines. At Wetwang Slack, we reconstruct a 13-generation pedigree comprising 195 individuals structured around female-line connections: matrilineal transmissions greatly outnumbered patrilineal ones and male reproductive partners were largely absent from the cemetery, plausibly because they were buried in their own natal communities. Furthermore, the three main sites with robust sample sizes were characterised by non-overlapping dominant mitochondrial haplogroups, implying a maternal clan-based structure. Reproductive unions at Wetwang Slack suggest a recurrent alliance between two dominant maternal descent groups, with members of each group never reproducing with members of their own maternal lineage. Meanwhile, three individuals from lavishly furnished chariot burials at Wetwang Slack were close maternal relatives belonging to a lineage with consecutive generations of close kin unions, a pattern largely absent among other individuals at the site. These results indicate highly distinctive social practices among an elite group embedded in the wider kinship network of the Arras community.

The earliest events during human tumor initiation, while poorly characterized, may hold clues to malignancy detection and prevention1. Here we model occult pre-neoplasia by bi-allelically inactivating TP53, a common early event in gastric cancer, in human gastric organoids. Causal relationships between this initiating genetic lesion and resulting phenotypes were established using experimental evolution in multiple clonally derived cultures over two years. TP53 loss elicited progressive aneuploidy, including copy number alterations and structural variants prevalent in gastric cancers, with evident preferred orders. Longitudinal single cell sequencing of TP53 deficient gastric organoids similarly indicates progression towards malignant transcriptional programs. Moreover, high-throughput lineage tracing with expressed cellular barcodes demonstrates reproducible dynamics whereby initially rare subclones with shared transcriptional programs repeatedly attain clonal dominance. This powerful platform for experimental evolution exposes stringent selection, clonal interference, and a striking degree of phenotypic convergence in pre-malignant epithelial organoids. These data imply predictability in the earliest stages of tumorigenesis and reveal evolutionary constraints and barriers to malignant transformation with implications for earlier detection and interception of aggressive, genome instable tumors.

The chromosomal theory of inheritance has dominated human genetics, including cancer genetics. Genes on the same chromosome segregate together while genes on different chromosomes assort independently, providing a fundamental tenet of Mendelian inheritance. Extrachromosomal DNA (ecDNA) is a frequent event in cancer that drives oncogene amplification, dysregulated gene expression and intratumoral heterogeneity, including through random segregation during cell division. Distinct ecDNA sequences, herein termed ecDNA species, can co-exist to facilitate intermolecular cooperation in cancer cells. However, how multiple ecDNA species within a tumor cell are assorted and maintained across somatic cell generations to drive cancer cell evolution is not known. Here we show that cooperative ecDNA species can be coordinately inherited through mitotic co-segregation. Imaging and single-cell analyses show that multiple ecDNAs encoding distinct oncogenes co-occur and are correlated in copy number in human cancer cells. EcDNA species are coordinately segregated asymmetrically during mitosis, resulting in daughter cells with simultaneous copy number gains in multiple ecDNA species prior to any selection. Computational modeling reveals the quantitative principles of ecDNA co-segregation and co-selection, predicting their observed distributions in cancer cells. Finally, we show that coordinated inheritance of ecDNAs enables co-amplification of specialized ecDNAs containing only enhancer elements and guides therapeutic strategies to jointly deplete cooperating ecDNA oncogenes. Coordinated inheritance of ecDNAs confers stability to oncogene cooperation and novel gene regulatory circuits, allowing winning combinations of epigenetic states to be transmitted across cell generations.

Sex differences in behaviour exist across all animals, typically under strong genetic regulation. In Drosophila, fruitless/doublesex transcription factors can identify dimorphic neurons but their organisation into functional circuits remains unclear. We present the connectome of the entire Drosophila male central nervous system. This contains 166,691 neurons spanning the brain and nerve cord, fully proofread and annotated including fruitless/doublesex expression and 11,691 types. We provide the first comprehensive comparison between male and female brain connectomes to synaptic resolution, finding 7,205 isomorphic, 114 dimorphic, 262 male-specific and 69 female-specific types. This resource enables analysis of full sensory-to-motor circuits underlying complex behaviours and the impact of dimorphic elements. Sex-specific/dimorphic neurons are concentrated in higher brain centres while the sensory and motor periphery are largely isomorphic. Within higher centres, male-specific connections are organised into hotspots defined by male-specific neurons or arbours. Numerous circuit switches reroute sensory information to form antagonistic circuits controlling opposing behaviours.

To learn effectively, animals must generalize across yet distinguish between related contexts. Generalization relies on low-dimensional neural manifolds found throughout neocortex1-3, which accelerate learning by constraining neural activity to task-relevant axes4,5. Conversely, context separation is attributed to neural expansion layers that can project information into high-dimensional feature spaces6-8, most famously cerebellar granule cells (GrCs)9-11. To investigate the generalization-separation tradeoff, we simultaneously imaged key nodes in the universal cortico-cerebellar pathway12,13--premotor layer 5 pyramidal tract (L5PT) and GrCs--during parallel learning of two distinct skills with shared temporal structure. Rather than expanding the cortical representations, GrCs retained their low-rank encoding of each task. Across contexts, however, despite stable cortico-cerebellar coupling, L5PT activity patterns generalized while GrC patterns temporally remapped. Mechanistically, GrCs used affine transformations that rotated the cortical manifolds apart but preserved their intrinsic low-dimensional geometry. Moreover, GrCs decorrelated cortical trajectories most strongly in expert animals. This reveals a fundamental architectural division of labor: the cortex generates invariant dynamic primitives for smooth generalization, while the cerebellum reconfigures them to drive context-specific output.

Bacterial, plant, and animal cells synthesize nucleotide immune signals as a conserved strategy to defend against viral infection1-4. In bacteria, Thoeris anti-phage defense systems convert nicotinamide adenine dinucleotide (NAD+) into the cyclic ADP-ribose signals 2'cADPR and 3'cADPR to activate downstream effectors and restrict viral replication5-8. Phage proteins can bind and sequester Thoeris signals6,9-13, but no mechanisms are known to degrade the exceptionally stable 2'cADPR and 3'cADPR molecules and terminate immune activation. Here we use a forward biochemical screen to discover the mycobacteriophage protein RyDEP as the founding member of an enzyme family that cleaves 2'cADPR and 3'cADPR to inactivate Thoeris defense. We show that RyDEP is a glycosidase that cleaves the ribose-ribose linkage in 2' and 3' cADPR immune signals to both inactivate host defense and enable direct restoration of NAD+. A crystal structure of the RyDEP-3'cADPR complex in the post-cleavage state explains the molecular basis of immune signal degradation and reveals surprising homology with the Repeat12 domain of animal ryanodine receptors (RyRs) that control calcium flux and muscle contraction14,15. We demonstrate that diverse phage RyDEP proteins tune RyR-domain activity to either degrade or sequester immune signals. Our results define RyR-domain proteins as regulators of nucleotide immune signaling and explain how viruses subvert host antiviral defense.

The gut microbiome is crucial for health and greatly affected by lifestyle. Many microbes common in non-industrialized populations are disappearing or extinct in industrialized populations. Understanding which microbes have been long-term residents of the human gut, and may have co-evolved with humans, has implications for the importance of microbial biodiversity loss for health. However, the genetic complexities of microbial evolution and the plasticity of gut microbiome composition have made it challenging to define these long-term associations. Here, we performed deep metagenomic sequencing of the Tsimane horticulturalists of Bolivia and compared their gut microbiomes with the Hadza hunter-gatherers of Tanzania. These two populations, whose ancestors were separated for tens of thousands of years, share 1,231 microbial species, most of which are absent in industrialized populations. Population genetic analyses in 636 of these shared species revealed patterns of microbial divergence and gene flow consistent with prehistoric human co-migration, with estimated split times that approximately align with human migration out of Africa and into the Americas. Our findings indicate that a diverse gut microbiome co-migrated with humans around the globe, persisting over millennia. However, many of these species are now vanishing in industrialized populations, and the consequences for human health remain uncertain.

Summary paragraphGABA ({psi}-aminobutyric acid) is the primary inhibitory neurotransmitter in the mammalian central nervous system (CNS) 1,2. There is a wide range of GABAergic neuronal types, each of which plays an important role in neural processing and the etiology of neurological disorders 3-5. However, there is no comprehensive understanding of this functional diversity, due to the lack of genetic tools to target and study the multitude of cell types. Here we perform two-photon imaging of GABA release in the inner plexiform layer (IPL) of the mouse retina using the newly developed GABA sensor iGABASnFR2. By applying varied light stimuli to isolated retinae, we reveal over 40 different GABA-releasing neurons, including some not previously described. Individual types show unique distributions of synaptic release sites in the sublayers comprising the IPL, allowing layer-specific visual encoding. Synaptic input and output sites are aligned along specific retinal orientations for multiple neuronal types. Furthermore, computational modeling reveals that the combination of cell type-specific spatial structure and unique release kinetics enables inhibitory neurons to suppress and sculpt excitatory signals in response to a wide range of behaviorally relevant motion structures. Our high-throughput approach provides the first comprehensive physiological characterization of inhibitory signaling in the vertebrate CNS. Future applications of this method will enable interrogation of the function and dysfunction of diverse inhibitory circuits in health and disease.

The entorhinal cortex (EC) is a highly-interconnected hub for multisensory integration and memory processing1-3, containing diverse neuronal subtypes4,5 including subpopulations that are uniquely spatially-tuned6,7. Although many spatial and memory functions develop in infancy, it is considered that neurogenesis and neuronal migration to the EC occurs prenatally. Here we show that the postnatal human temporal lobe contains a prominent stream with large chains of young migrating neurons and many individual neurons breaking away directed into the EC. The EC stream forms between the second and third trimesters of prenatal development when the lateral ventricle walls in the temporal lobe collapse, displacing the subventricular zone (SVZ) and dividing radial glia. At birth, the EC stream follows a path of radial glial [fi]bers in the site of the collapsed ventricle. Migratory chains persist up to 11 months postnatally; however, many individually migrating young neurons can still be detected in the EC at 2 years of age and a few isolated cells at 3 years of age. Within the EC at birth, immature neurons are a mixed population expressing markers of the medial ganglionic eminence (MGE) and caudal ganglionic eminence (CGE), but postnatally rapidly become primarily CGE-derived. Using single-nuclei RNAseq we identified these lineages and found that the MGE-derived neurons matured at earlier postnatal ages compared to those derived from the CGE. The CGE interneurons arriving and maturing the latest included subtypes expressing calretinin (CR), reelin (RELN), and vasoactive intestinal protein (VIP) many of which settle in layer II of the entorhinal cortex. This study reveals that the human EC is still being constructed during the first years of life revealing the largest known postnatal stream of migratory neurons in humans. The protracted postnatal arrival of a diverse population of interneurons could contribute to plasticity8,9 and proper excitation-inhibition balance10,11 within these highly connected brain circuits.