Nature

In Pang et al. (2023)1, we identified a close link between the geometry and function of the human brain by showing that: (1) eigenmodes derived from cortical geometry parsimoniously reconstruct activity patterns recorded with functional magnetic resonance imaging (fMRI); (2) task-evoked cortical activity results from excitations of brain-wide modes with long wavelengths; (3) wave dynamics, constrained by geometry and distance-dependent connectivity, can account for diverse aspects of spontaneous and evoked brain activity; and (4) geometry and function are strongly coupled in the subcortex. Faskowitz et al. (2023)2 raise concerns about the framing of our paper and the specificity of the eigenmode reconstructions in result (1). Here, we address these concerns and show how specificity is established by using appropriate benchmarks.

Orphan G protein-coupled receptors (GPCRs) are largely intractable therapeutic targets, owing to the lack of chemical tools for exploring their pharmacology. The discovery of such tools, however, is hampered by a number of unknowns, such as effector coupling and appropriate positive controls. In our 2017 Nature Chemical Biology paper1, we developed a computational chemical tool discovery approach called GPCR Contact-Informed Neighboring Pocket (GPCR-CoINPocket). This method predicted pharmacological similarity of GPCRs in a ligand- and structure-independent manner, to enable the discovery of off-target activities of known compounds at orphan GPCRs and hence the identification of so-called surrogate ligands. Our orphan GPCR target for prospective surrogate ligand discovery efforts was GPR37L1, a brain-specific receptor linked to cerebellar development2 and seizures3. We had previously demonstrated that GPR37L1 constitutively coupled to Gs and generated ligand-independent increases in intracellular cAMP4[§]. Thus, the inverse agonist activities of computationally predicted surrogates were tested in the cAMP response element luciferase (CRE-luc) reporter gene assay in human embryonic kidney (HEK293) cells expressing either vector control or what we thought was untagged GPR37L1 in pcDNA3.1. However, we recently discovered that the GPR37L1 construct used in that study was incorrect: instead of pcDNA3.1, it carried the receptor inserted backwards into a yeast p426GPD vector (hereafter referred to as p426-r37L1). Here, we correct the cloning error and describe our subsequent unsuccessful efforts to re-test the computationally predicted GPR37L1 ligands (triggering an author-initiated retraction of1). NoteWe, the authors, are working with the Nature Chemical Biology Editors to retract our 2017 paper Orphan receptor ligand discovery by pickpocketing pharmacological neighbors1. The present manuscript is under review at Nature Chemical Biology as a Matters Arising accompaniment to the anticipated author-initiated retraction. We initiated the steps towards the retraction upon discovering a regrettable cloning error that put into question the in vitro findings reported in1. This action was unanimously agreed upon by all authors. The computational aspects of the original manuscript1 are unaffected by this error.

Critical illness in COVID-19 is caused by inflammatory lung injury, mediated by the host immune system. We and others have shown that host genetic variation influences the development of illness requiring critical care1 or hospitalisation2;3;4 following SARS-Co-V2 infection. The GenOMICC (Genetics of Mortality in Critical Care) study recruits critically-ill cases and compares their genomes with population controls in order to find underlying disease mechanisms. Here, we use whole genome sequencing and statistical fine mapping in 7,491 critically-ill cases compared with 48,400 population controls to discover and replicate 22 independent variants that significantly predispose to life-threatening COVID-19. We identify 15 new independent associations with critical COVID-19, including variants within genes involved in interferon signalling (IL10RB, PLSCR1), leucocyte differentiation (BCL11A), and blood type antigen secretor status (FUT2). Using transcriptome-wide association and colocalisation to infer the effect of gene expression on disease severity, we find evidence implicating expression of multiple genes, including reduced expression of a membrane flippase (ATP11A), and increased mucin expression (MUC1), in critical disease. We show that comparison between critically-ill cases and population controls is highly efficient for genetic association analysis and enables detection of therapeutically-relevant mechanisms of disease. Therapeutic predictions arising from these findings require testing in clinical trials.

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.

The SARS-CoV-2 pandemic has led to an urgent need to understand the molecular basis for immune recognition of SARS-CoV-2 spike (S) glycoprotein antigenic sites. To define the genetic and structural basis for SARS-CoV-2 neutralization, we determined the structures of two human monoclonal antibodies COV2-2196 and COV2-21301, which form the basis of the investigational antibody cocktail AZD7442, in complex with the receptor binding domain (RBD) of SARS-CoV-2. COV2-2196 forms an "aromatic cage" at the heavy/light chain interface using germline-encoded residues in complementarity determining regions (CDRs) 2 and 3 of the heavy chain and CDRs 1 and 3 of the light chain. These structural features explain why highly similar antibodies (public clonotypes) have been isolated from multiple individuals1-4. The structure of COV2-2130 reveals that an unusually long LCDR1 and HCDR3 make interactions with the opposite face of the RBD from that of COV2-2196. Using deep mutational scanning and neutralization escape selection experiments, we comprehensively mapped the critical residues of both antibodies and identified positions of concern for possible viral escape. Nonetheless, both COV2-2196 and COV2-2130 showed strong neutralizing activity against SARS-CoV-2 strain with recent variations of concern including E484K, N501Y, and D614G substitutions. These studies reveal germline-encoded antibody features enabling recognition of the RBD and demonstrate the activity of a cocktail like AZD7442 in preventing escape from emerging variant viruses.

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.

The ability to act voluntarily is fundamental to animal behavior1,2,3,4,5. For example, self-directed movements are critical to exploration, particularly in the absence of external sensory signals that could shape a trajectory. However, how neural networks might plan future changes in direction in the absence of salient sensory cues is unknown. Here we use volumetric two-photon imaging to map neural activity associated with walking across the entire brain of the fruit fly Drosophila, register these signals across animals with micron precision, and generate a dataset of [~]20 billion neural measurements across thousands of bouts of voluntary movements. We define spatially clustered neural signals selectively associated with changes in forward and angular velocity, and reveal that turning is associated with widespread asymmetric activity between brain hemispheres. Strikingly, this asymmetry in interhemispheric dynamics emerges more than 10 seconds before a turn within a specific brain region associated with motor control, the Inferior Posterior Slope (IPS). This early, local difference in neural activity predicts the direction of future turns on a trial-by-trial basis, revealing long-term motor planning. As the direction of each turn is neither trained, nor guided by external sensory cues, it must be internally determined. We therefore propose that this pre-motor center contains a neural substrate of volitional action.

Summary AbstractSpatially integrated mechanisms of consciousness are unclear1,2. An approach to manipulate brainwide circuits regulating consciousness via synthetic central nervous system activation may pave the way for more precise transitions in consciousness and reveal underlying mechanisms. Toward this goal, we leverage anesthesia as a tool to probe consciousness at cellular resolution within the intact network. We perform brainwide chemogenetic capture3,4 of isoflurane anesthesia-activated circuitry in mice --in parallel with electrocorticography5, wireless mechano- acoustic recording of peripheral physiology6, and behavioral classification7,8-- to describe a synthetic state of altered consciousness generated in the absence of an anesthetic agent. We define patterns of activation under isoflurane using intact brain immediate early gene mapping9-12 combined with brainwide high density silicon probe recordings13. Our data identify subcortical hotspots of neural activity in an unconsciousness network that is globally characterized by increased functional connectivity driven by select nodes. We provide technical resources spanning brainwide single-cell resolution maps and neurophysiologic datasets of the isoflurane-rendered unconscious state, along with an approach to further probe its global cellular-level mechanisms. Together, we present the foundation for future research to refine this viral-genetic brainwide approach to generate synthetic conscious state transitions, such as sleep, stasis, analgesia or anesthesia.

The woolly rhinoceros (Coelodonta antiquitatis) is an iconic species of the Eurasian Pleistocene megafauna, which was abundant in Eurasia in the Pleistocene until its demise beginning approximately 10,000 years ago. Despite the early recovery of several specimens from well-known European archaeological sites, including its type specimen (Blumenbach 1799), no genomes of European populations were available so far, and all available genomic data originated exclusively from the Siberian population1. Using coprolites of cave hyenas (Crocuta crocuta spelea) recovered from Middle Palaeolithic layers of two caves in Germany (Bockstein-Loch and Hohlenstein-Stadel), we isolated and enriched predator and prey DNA to assemble the first European woolly rhinoceros mitogenomes, in addition to cave hyena mitogenomes. These mitogenomes of European woolly rhinoceros are genetically distinct from the Siberian woolly rhinoceros, and analyses of the more complete mitogenome suggests a split of the populations potentially coinciding with the earliest fossil records of wooly rhinoceros in Europe.

Developmental variation in brain-wiring contributes to behavioural individuality1,2. However, how and when individualized wiring diagrams emerge and become stable during development remains largely unknown. Here, we explored axon targeting dynamics in individual brains using live-imaging of a developing Drosophila visual circuit and discovered that targeting choice is an algorithmic multi-step growth process with variable outcomes. Using optogenetics, we found that temporally restricted Notch lateral-inhibition defines a subset of neurons with a probabilistic potential to innervate distal targets. Next, axons from NotchOFF neurons amplify into long actin-rich multi-fibre structures necessary for distal growth. A subset of these NotchOFF neurons create distal targeting axons by stabilizing microtubule growth in one of their actin fibres. Amplified axons without tubulin-stabilized fibres retract, resulting in the stochastic selection of a different number of distal targeting axons in each brain. Pharmacological microtubule destabilization suffices to inhibit this targeting. We observed a similar axonal amplification-stabilization process in the developing chick spinal cord, suggesting a conserved mechanism. Finally, early microtubule patterns predict the adult brain-wiring of an individual in a target-independent manner prior to synapse formation3,4. Thus, we show that a temporal succession of genetically encoded stochastic processes explains the emergence of individual wiring variation. One-Sentence SummaryThe temporal succession of stochastic developmental processes explains the emergence of individual wiring variation.

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.

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 emergence of eukaryotes from a merger between an archaeon and a bacterial cell [~]two billion years ago involved a profound change in cellular organisation. While the order in which different features of the eukaryotic cell arose remains a matter of controversy1-3, close archaeal relatives of eukaryotes have recently been identified that possess homologues of eukaryotic trafficking machinery4,5 and a complex cell architecture6,7. The members of this phylum, the Asgard archaea (syn. Prometheoarchaeota) described so far, however, lack internal membrane-bound compartments, and therefore have shed little light on origins of the hallmark eukaryotic endomembrane system. Here we report the cell biological analysis of a member of the Heimdallarchaeia, Candidatus Y. umbracryptum in enriched mixed microbial communities. Possessing a small genome encoding few homologues of eukaryotic membrane remodelling machinery8, Ca. Y. umbracryptum cells in late-stage cultures resemble previously described Asgard archaea with extensive cellular protrusions. Surprisingly, during early stages of culture growth Ca. Y. umbracryptum cells have fewer protrusions but possess numerous intracellular vesicles, most of which have a luminal surface that morphologically resembles the outer coat of the plasma membrane. These data alter our view of eukaryogenesis by identifying a close archaeal relative of eukaryotes with an endomembrane system, whose appearance changes with cell state.

Replying to GCC Koh et al., Nature Genetics (www.nature.com/articles/s41588-023-01602-9) and the comments by Simon Boulton in Nature Genetics (www.nature.com/articles/s41588-023-01611-8). The drug CX-5461 [Pidnarulex] is an inhibitor of RNA polymerase I and topoisomerase II and a DNA G-quadruplex stabilizer. Koh et al. recently reported that CX-5461 induces extensive, non-selective collateral mutagenesis in vitro at magnitudes surpassing known environmental carcinogens, raising concerns about its potential to promote secondary cancers. Mindful that the report of Koh et al. was exclusively an in vitro study, we applied the same ultra-sensitive, error-corrected TwinStrand Duplex Sequencing approach and analyses (methods) to a longitudinal series of clinical specimens from four patients who participated in the Phase I dose-escalation trial of CX-5461 in advanced haematologic malignancies [ACTRN12613001061729]. In contrast to the findings of Koh et al. we found no evidence that clinical administration of CX-5461 significantly increased mutational burden in patient samples, nor did we observe the reported mutational signature. These results suggest that the mutagenic effects described in vitro do not translate to the clinical setting. This is particularly important given that several clinical trials, including a Phase Ib study of CX-5461 in patients with solid tumours [NCT04890613] and a National Cancer Institute sponsored Phase I trial [NCT06606990] are currently enrolling patients. These ongoing and planned clinical efforts highlight both the therapeutic promise of CX-5461 and the importance of evaluating this agent within rigorous and clinically relevant frameworks.

HIV persists in long-lived CD4 T cell reservoirs despite antiretroviral therapy (ART)1. Elite controllers suppress viraemia without ART2, yet reservoir reactivation emerges with immune ageing3. Here we show that transient reprogramming of patient-derived CD4 T cells restores their ability to eliminate HIV-infected reservoirs, excising integrated proviral DNA. A defined compound, or physiological induction, drove rapid reprogramming ex vivo, enabling clearance of HIV DNA within hours to days of treatment, independently of ART. Single-cell RNA sequencing revealed activation of an antiviral telomere transfer programme4 that exceeds elite-like control. In humanised mice, adoptive transfer of reprogrammed patient CD4 T cells, or in vivo reprogramming of murine T cells, eliminated HIV DNA across reservoirs, with undetectable viral genomes persisting for months. Modelling predicted that residual proviral reactivation would be governed by rare stochastic events, rendering viral rebound unlikely within a human lifespan. These findings identify a previously unrecognised form of intracellular immunity and establish a defined route to a functional HIV cure arising from the CD4 T cell itself.

Mitochondria and nucleotide metabolism are critical for cellular and developmental homeostasis, yet their potential interdependence and role in neurodevelopmental disease remain unclear. In MECP2 Duplication Syndrome (MDS), we identify a conserved correlation between mitochondrial function and purine metabolism that is disrupted across human, organoid, and mouse models. Multiomics integration reveals Complex III as the focal point of mitochondrial collapse, leading to redox stress, DNA damage, and hyperactivation of the de novo purine biosynthesis via purinosome assembly. The breakdown of mitochondria-purinosome coupling compromises genome stability, impairs radial glia proliferation, and delays neuronal maturation. By linking a defined genetic dosage imbalance to metabolic network failure, our study positions the mitochondria-purinosome coordination as a fundamental control axis for neurodevelopment and a therapeutic entry point across metabolic and neurodevelopmental disorders. Metabolic control is fundamental to cellular function, influencing energy production, signaling, epigenetic regulation, and tissue homeostasis1. Nowhere is this more critical than in the brain, where tightly regulated metabolic networks sustain high energetic demands and support neuronal development, synaptic plasticity, and circuit formation2. Disruptions in these networks are increasingly implicated across a spectrum of neurodevelopmental disorders3-5, yet their precise metabolic signatures and mechanistic contributions remain poorly understood.

Our understanding of the neurobiology of primate behavior largely derives from artificial tasks in highly-controlled laboratory settings, overlooking most natural behaviors primate brains evolved to produce1-3. In particular, how primates navigate the multidimensional social relationships that structure daily life4 and shape survival and reproductive success5 remains largely unexplored at the single neuron level. Here, we combine ethological analysis with new wireless recording technologies to uncover neural signatures of natural behavior in unrestrained, socially interacting pairs of rhesus macaques. Single neuron and population activity in prefrontal and temporal cortex unveiled robust encoding of 24 species-typical behaviors, which was strongly modulated by the presence and identity of surrounding monkeys. Male-female partners demonstrated near-perfect reciprocity in grooming, a key behavioral mechanism supporting friendships and alliances6, and neural activity maintained a running account of these social investments. When confronted with an aggressive intruder, behavioral and neural population responses reflected empathy and were buffered by the presence of a partner. By employing an ethological approach to the study of primate neurobiology, we reveal a highly-distributed neurophysiological ledger of social dynamics, a potential computational foundation supporting communal life in primate societies, including our own.

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.

Pediatric trials are ethically and logistically difficult, so the U.S. FDA often extrapolates adult data to children when justified. Yet no public resource systematically documents these decisions. We present PED-X-Bench, the first dataset and benchmark that encodes FDA pediatric-extrapolation outcomes as a four-way classification task (Full, Partial, None, Unlabeled). PED-X-Bench contains 737 FDA drug-label sections ({approx} 1 M words of source text) for approvals issued 2007-2024 across all therapeutic areas. A two-stage o3-mini prompting pipeline mined full FDA label text; nine domain reviewers then adjudicated a stratified sample of 135 labels yielding an accuracy F1 of 0.74 and 0.63 respectively (inter-annotator {kappa} = 0.678) and spot-checking the remainder. For every drug we release the ground-truth label, concise efficacy and pharmacokinetic/safety summaries, and harmonized study metadata. To showcase utility we release two baseline models: (i) a logistic-regression classifier that uses structured metadata from FDAs pediatric-labeling dataset, and (ii) a fine-tuned BigBird BERT that ingests full label text. Both base-lines perform modestly, leaving ample headroom for future work. PED-X-Bench enables research on pediatric drug development, clinical NLP and drug safety; dataset card and code are made available here: github.com/tatonetti-lab/PedXBench huggingface.co/datasets/apoorvasrinivasan/Ped-X-Bench

Animals learn about the external world, in part, via interoceptive signals1,2. For example, the nutrient content of food is first estimated in the mouth, in the form of flavor, and then measured again via slower signals from the gut. How these signals from the mouth and gut are integrated to drive learning is unknown. Here we identify a lateralized dopamine pathway that is specialized for learning about the nutrient content of food. We show that dopamine neurons in the ventral tegmental area (VTADA) are necessary for associating nutrients with flavors, and that post-ingestive nutrients trigger DA release selectively in a small region of the anterior basolateral amygdala (BLA) but not canonical DA targets in striatum. Remarkably, this nutrient-triggered DA release occurs preferentially on the left side of the brain in both mice and humans, revealing that the DA system is functionally lateralized. We identify the gut sensors that are responsible for nutrient-triggered DA release; show that they activate BLA-projecting DA neurons defined by expression of cholecystokinin (CCK); and demonstrate that stimulation of DA axon terminals in the anterior BLA drives flavor-nutrient learning but not other aspects of feeding behavior. Two-photon imaging of neurons in the left anterior BLA reveals that they integrate gustatory and post-ingestive cues, and silencing these neurons prevents flavor-nutrient learning. These findings establish a neural basis for how animals learn about the nutrient content of their food. They also reveal unexpectedly that post-ingestive nutrients are differentially represented on the right and left sides of the brain.