Mechanical pain that resists even systemic opioids is a major clinical challenge, yet its underlying mechanisms remain poorly understood. Here, we identify a glycinergic descending circuit from GlyT2-positive neurons in the rostral ventromedial medulla (RVMGlyT2+) to somatostatin-positive neurons in the spinal dorsal horn (SDHSOM+). This RVMGlyT2+→SDHSOM+ circuit is both necessary and sufficient for widespread morphine-resistant mechanical allodynia in mice. Craniofacial pain recruits this circuit through trigeminal and limbic inputs. Mechanistically, glycine from RVMGlyT2+ terminals activates GluN3A-containing NMDA receptors on SDHSOM+ neurons, where they function as unconventional excitatory glycine receptors. This "Glycine-GluN3A" signaling bypasses opioid suppression, unmasks latent mechanical allodynia, and is required for the opioid-resistant pain state. Our findings uncover an unexpected excitatory glycinergic mechanism in the spinal cord, define a dedicated circuit for morphine-resistant pain, and highlight GluN3A-containing NMDARs as potential therapeutic targets for intractable pain.
Tau pathology closely tracks neuronal loss in Alzheimer's disease, but how phosphorylated tau becomes lethal has remained unclear. Chen et al. identify a dual-hit mechanism: glucose hypometabolism removes a protective A20-mediated brake on necroptosis while phosphorylated tau scaffolds RIPK1 activation, driving tau-associated neuronal death.1.
The sleeping brain processes memory-related semantic information, yet decoding such content remains challenging. Combining a large, openly shared sleep EEG dataset with neural contrastive learning, Chen et al.,1 in this issue of Neuron, establish a promising framework for non-invasive decoding of the sleeping human mind.
When humans and other animals assess that they have little control over outcomes, their motivation and capacity for learning change. However, how we first learn what the control level is and how it then influences learning remain unclear. Using functional magnetic resonance imaging (fMRI) and transcranial magnetic stimulation (TMS) in human participants, we provide both recording and transient disruption evidence for an anatomically grounded model of control learning, in which dorsomedial prefrontal cortex (dmPFC) tracks confidence in each individual decision in order to then estimate controllability and changes in controllability via interactions with the dorsal raphe nucleus. Disrupting activity in this circuit, via dmPFC manipulation, impaired control learning. In addition, another mechanism was identified through which control estimates influenced learning. Participants’ estimates of controllability were associated with widespread changes in outcome signals, including in dopamine-associated midbrain nuclei and in credit-assignment-linked signals in a distinct prefrontal cortex subregion.
Large language models (LLMs) have mastered human language in ways that no previous computational system has. While rule-based, symbolic systems sufficed for constrained, well-defined problems, they were not able to accommodate the context-sensitive expressivity of natural language. LLMs instead use statistical learning to encode the diversity of linguistic structures into a unified high-dimensional embedding space. Strikingly, this context-driven, distributed representation closely parallels neural population codes, suggesting that the human language system may have converged on a similar computational strategy. Drawing on a growing body of work at the intersection of artificial intelligence and cognitive neuroscience, we show that LLMs can serve as cognitively plausible models of the neural computations supporting language in the human brain. We conclude that explaining how language can emerge from neural population codes, in both biological and artificial systems, will not be achieved through the incremental refinement of algebraic-symbolic theories but will demand new theoretical paradigms.
Astrocytes support synaptic activity and associated circuit function, including those linked to sleep and memory, though how they coordinate these processes remains poorly defined. Here, we show that the transcription factor Nuclear Factor I-X (NFIX) is required to maintain astrocyte function in the thalamic reticular nucleus (TRN), as its deletion results in aberrant thalamocortical oscillations and impaired memory formation. We found that NFIX directly regulates the expression of monoamine oxidase B (MAOB) and the purinergic receptor, P2RX7. Mechanistically, knockdown of Maob or knockout of P2rx7 in TRN astrocytes results in decreased astrocytic release of the neurotransmitter GABA, reducing tonic inhibition of ventrobasal (VB) relay neurons and impairing memory formation. Altogether, our studies identify region-specific roles for astrocytic NFIX and its gene regulatory network in TRN astrocytes, while further revealing that astrocytes in the TRN couple regulation of thalamocortical circuit dynamics and memory formation through parallel GABA and purinergic signaling pathways.
Aging is the major risk factor for neurodegenerative disease, yet the mechanisms linking physiological aging to brain dysfunction remain unclear. We investigated the brains of telomere-shortened mice and observed lipofuscinosis, hypomyelination, microglial atrophy, and cognitive deficits. Single-nucleus RNA sequencing (snRNA-seq) revealed accelerated glial aging and elevated microglial senescence pathways. In a senescence model of human induced pluripotent stem cell (iPSC)-derived microglia, delta-like non-canonical Notch ligand 1 (DLK1) was identified as a novel senescence-associated ligand. Soluble DLK1 (sDLK1) was increased in the cerebrospinal fluid of telomere-shortened and naturally aged mice, and this increase was eliminated by microglial depletion. In vivo elevation of sDLK1 caused hypomyelination and blocked oligodendrocyte lineage progression, and these effects demonstrate the detrimental nature of excessive sDLK1. In human iPSC systems, sDLK1 impaired oligodendrocyte maturation and altered calcium signaling in excitatory neurons. These findings identify microglial senescence as a core consequence of telomere shortening and reveal sDLK1 as a microglia-derived senescence ligand that drives oligodendrocyte and neuronal dysfunction in aging.
The decision to engage in competition, enabling individuals to acquire resources with minimal risk and energetic cost, is based on prior outcomes. Yet the neural mechanism underlying this decision-making process remains unclear. Here, we developed a win maze, allowing mice to voluntarily decide whether to re-engage in competition following an outcome. Dominant males showed a stronger preference for competition engagement than subordinate males. Mechanistically, dorsomedial prefrontal cortex (dmPFC) population activity exhibited rotational dynamics during the task, with the competition process driving an enlarged loop trajectory, and remained within the competition-related latent space during consecutive win trials. Meanwhile, winning outcomes evoked dmPFC dopamine release, which was associated with competition re-engagement. Consistently, inhibiting the ventral tegmental area (VTA)DA → dmPFC pathway or dmPFC dopamine D1 receptor-positive (D1R+) neurons during winning outcomes abolished re-engagement in dominant males, whereas activating the VTADA → dmPFC pathway enhanced engagement in subordinate males. Together, these findings reveal a status-dependent dopamine mechanism that translates competitive outcomes into future engagement decisions.
Hippocampal circuits are composed of cells that exhibit heterogeneity in gene expression, connectivity, and intrinsic properties. This diversity arises during embryonic development, which produces parallel subcircuits along the trisynaptic pathway. To examine how these subcircuits support spatial memory, we combined embryonic birthdating of interconnected CA3-CA1 neurons with electrophysiology in a place-reward association task. Learning reorganized correlation patterns near rewarded locations, which were reactivated during sleep and predicted memory retrieval. Neurons born on the same day exhibited correlated activity across brain states and hippocampal subfields. Throughout learning, same-birthdate CA1 neurons coordinated their activity to encode the rewarded locations, but CA3 cells did not. These regional differences were mirrored by distinct patterns of connectivity between same-birthdate pyramidal cells and inhibitory interneurons. In particular, same-birthdate neurons converged onto parvalbumin-expressing CA1 interneurons, which were modulated by spatial learning. Together, our results demonstrate that development-defined subcircuits in the hippocampus are preconfigured to encode new memories.
Hyperphosphorylation and aggregation of tau are pathological hallmarks of tauopathies. Mitochondrial dysfunction is also a common feature of tauopathies. The mechanistic link between tau abnormalities and mitochondrial dysfunction and its relationship to the physiological function of tau, however, is unclear. Here, we demonstrate that tau regulates mitochondrial reverse electron transport (RET), which produces excess reactive oxygen species (ROS), reduces the NAD+/NADH ratio, and is activated by aging or stress. In flies, mice, and human induced pluripotent stem cell (hiPSC)-derived neurons, tau depletion eliminates stress-induced RET and confers resilience. Mechanistically, tau enters mitochondria and directly interacts with the complex I subunit NDUFS3 to promote RET in a phosphorylation-dependent manner. Elevated RET further drives tau hyperphosphorylation, establishing a self-perpetuating pathological loop. Inhibition of RET ameliorates tau toxicity across species. RET regulation thus represents a previously unrecognized normal function of tau that becomes pathological in disease, providing a therapeutic target for various conditions characterized by tau abnormalities and mitochondrial dysfunction.
Protein misfolding and propagation contribute to neurodegenerative diseases. Recently, cryogenic electron microscopy of insoluble amyloid fibrils derived from individuals with frontotemporal lobar degeneration (FTLD) revealed new species of amyloid fibrils, which are composed of aggregated TATA-binding protein-associated factor 15 (TAF15). However, it remains unknown whether TAF15 fibrils propagate in a prion-like manner and drive neurodegeneration. Here, we show that TAF15 forms amyloid fibrils that can self-propagate. Strikingly, a single injection of synthetic TAF15 pre-formed fibrils into the prefrontal cortex of wild-type mice led to the aggregation of endogenous TAF15 and cell-to-cell transmission of pathologic TAF15. TAF15 pathology was accompanied by progressive degeneration of cortical neurons, cognitive impairments, and anxiety- and depression-like behaviors. The detrimental effects of TAF15 fibrils were abolished by genetic deletion of endogenous TAF15. Together, these observations indicate that TAF15 aggregation drives neurodegeneration.
For decades, the cerebellar cortex was viewed as a simple circuit comprised of repeated modules containing only a few basic cell types. However, discoveries made possible by modern molecular and physiological approaches have challenged this traditional view. In particular, single-nucleus RNA sequencing (snRNA-seq) revealed that every major class of cerebellar neuron consists of multiple transcriptionally distinct subtypes. Here, we explore how mapping and functionally characterizing this transcriptomic diversity are fundamentally rewriting our understanding of neural computation in the cerebellum. We focus initially on molecular layer interneurons (MLIs), which exemplify the power of this approach: transcriptomics divides MLIs into distinct subtypes, which were subsequently discovered to perform entirely opposing computational roles, dictating Purkinje cell firing and regulating dendritic calcium signals critical for synaptic plasticity and learning. This same multi-modal "playbook" can now be leveraged to determine how subtypes of granule cells, Golgi cells, Purkinje layer interneurons, Purkinje cells, and unipolar brush cells are specialized to meet distinct computational needs that allow the cerebellum to contribute to a wide range of behaviors. Ultimately, the cellular diversity unmasked in the cerebellum offers a uniquely tractable model for learning the rules by which molecular diversification of cell types equips brain circuits with the computational flexibility needed to drive complex behaviors.
Behavior arises from the coordinated activity across anatomically and functionally distinct brain regions. Modern experimental tools allow unprecedented access to large neural populations spanning many interacting regions brain-wide. Yet, understanding such large-scale datasets necessitates robust, scalable computational models to extract meaningful features of inter-region communication and principled theories to interpret those features. Here, we introduce current-based decomposition (CURBD), an approach for inferring brain-wide interactions using data-constrained recurrent neural network models that autonomously produce dynamics consistent with experimentally obtained neural data. CURBD leverages the functional interactions inferred from such models to reveal directional currents between multiple brain regions simultaneously. We first show that CURBD accurately isolates inter-region currents in simulated, ground-truth networks with known connectivity and dynamics. We then apply CURBD to multi-region neural recordings obtained from many species-larval zebrafish, mice, macaques, and humans-to demonstrate the widespread applicability of CURBD in untangling brain-wide interactions and inter-area communication principles underlying behavior.
Impaired oligodendrocyte precursor cell (OPC) differentiation limits myelin renewal in aging and contributes to multiple sclerosis (MS) progression. How aging drives OPC deficits remains incompletely understood. We find dysregulation of genes associated with the circadian clock, including Bmal1, and metabolism in aged compared with young OPCs. Targeted loss of Bmal1 in OPCs drives metabolic dysfunction, leading to cellular senescence and impaired dynamics. OPC proliferation and differentiation occur at different rates throughout the day in young adult mice and become disrupted with aging. Chronotherapeutic targeting of BMAL1-controlled sirtuin signaling restores Bmal1-disrupted OPC dynamics after demyelination via sirtuin 2 (Sirt2)-dependent mechanisms. Induced pluripotent stem cell (iPSC)-derived OPCs from MS patients and MS lesion oligodendroglia recapitulate BMAL1 and SIRT2 disruptions. These findings establish BMAL1 as a key regulator of OPC energy metabolism, sirtuin homeostasis, and senescence. We anticipate that this work will provide a foundation for future studies investigating the interconnected roles of aging, circadian disruption, and myelin biology.
Early-life stress increases gene expression, neurophysiological, and behavioral responses to subsequent stress. Here, we determined the role of chromatin in such long-lasting sensitivity. We used a combination of bottom-up mass spectrometry, viral-mediated epigenome editing, RNA sequencing, patch-clamp electrophysiology of dopamine neurons, and behavioral quantification in a mouse model of early-life stress, focusing on the ventral tegmental area (VTA), a key dopaminergic brain region. We found that early-life stress enriches histone-3 lysine-4 monomethylation-associated with open chromatin and primed or active enhancers-and the H3K4 monomethylase SETD7. Mimicking early-life stress through postnatal overexpression of Setd7 and enrichment of H3K4me1 in the VTA sensitizes transcriptional, physiological, and behavioral responses to adult stress, while Setd7 knockdown ameliorates the impact of early-life stress. These findings link early-life stress experience to long-term stress hypersensitivity within the brain's dopaminergic circuitry, providing a mechanism by which early-life stress increases risk for mood and anxiety disorders later in life.
First discovered in anesthetized animals, neural traveling waves (nTWs) have now been observed throughout the brains of awake animals, where they influence neural excitability and behavior. nTWs can arise intrinsically in ongoing network activity or be triggered by sensory input and behavioral events. By structuring neural activity within individual cortical regions, nTWs introduce spatiotemporal dependencies across sensory maps that are not naturally captured by purely feedforward or feedback processing. Here, we synthesize physiological and computational evidence around two themes, focusing on the visual system while drawing connections to other cortical areas. First, we define nTWs and outline the circuit mechanisms that can generate them. Second, we highlight how nTWs can implement spatiotemporal computations, such as predicting upcoming sensory inputs, by embedding sensory history in the evolving activity pattern of an individual cortical region. We conclude by introducing a conceptual framework for spatiotemporal, generative processing by nTWs traveling over sensory maps.
The goal of artificial intelligence (AI) modeling in neuroscience, or NeuroAI, is to uncover the factors that give rise to human-level intelligence. However, current models overwhelmingly focus on simulating adulthood, the end state of intelligence, and often do not consider how this state was achieved in the first place. We argue that, to understand adult intelligence, it is important to model the developmental process by which intelligence arose. Here, we describe how developmental changes in children’s neural architecture, experiences, and learning objectives are adaptively suited to support rapid learning and illustrate how these principles can be incorporated into the AI engineering process. Finally, we describe the early developing capacities of children and how these capacities provide an ideal set of benchmarks for evaluating AI models. Together, by modeling the developmental process by which humans achieve intelligence, we may build more mechanistically plausible models as well as improve the capabilities of AI.
Gastrointestinal nutrient signals, arriving minutes after eating, powerfully reinforce food intake. How the brain links these delayed interoceptive signals to immediate sensory cues remains unclear. We show that ventral tegmental area (VTA) dopamine neurons resolve this credit-assignment problem by entering a distinct network state characterized by synchronized ∼0.8 Hz bursting, extending the integration window for learning across the gut-brain axis. Using two-photon imaging, Neuropixels recordings, and in vivo dopamine sensors in behaving mice, we identify this “sync state” emerging tens of seconds after ingestion begins, when orosensation overlaps with post-ingestive nutrient signals. This activity cannot be explained by reward prediction error or motor signals but predicts learning-dependent enhancement of future cue-evoked consumption vigor. Silencing dopamine neurons during oral-gastric overlap blocks nutrient learning without affecting intake, while optogenetic stimulation mimicking synchrony is sufficient to drive learning without nutrients. These findings reveal a novel dopaminergic mechanism for interoceptive credit assignment.
