Neuron

Decisions often involve trade-offs between benefits and costs, such as a small payment now vs. a larger payment after a delay. Although decision variables are distributed across frontal cortex, how these regions transform competing attributes into a choice remains unclear. Combining large-scale electrophysiology with causal perturbations as mice chose between different reward amounts and delays, we found that population geometry of decision variables differed across regions, reflecting distinct computations. Dorsal frontal areas, including motor cortex, collapsed competing attributes onto a decision axis shared across reward-delay contexts, with projection amplitude predicting choice probability; silencing these areas impaired action selection regardless of delay. In contrast, ventral prefrontal cortex (vPFC) represented choice in higher-dimensional subspaces organized by reward delay, whose geometry predicted how strongly delay influenced choice. Notably, inhibiting vPFC preserved the dorsal decision representation but selectively removed its modulation by reward delay, causing animals to behave as if delay no longer influenced choice. vPFC inhibition had no effect when mice chose between reward amounts. Frontal cortex therefore solves multi-attribute decisions through a generator-modulator architecture: dorsal frontal circuits maintain a shared choice axis guiding decisions across contexts, whereas vPFC maintains competing attributes in distinct subspaces that modulate decisions when options conflict.

Phasic ensemble bursting in subcortical neuromodulatory systems such as the basal forebrain (BF) provides a powerful format for broadcasting reward-prediction signals. These bursts encode reward prediction by varying population amplitude, yet their internal structure, reflected in cell-to-cell amplitude relationships, remains conserved. How neural circuits reconcile these competing constraints remains unknown. Here we show that BF bursting is generated by a dynamical attractor: burst structure is imposed not by feedforward inputs but by recurrent circuit dynamics, while burst amplitude is determined by approximately linear integration across concurrent inputs. Optogenetic activation of BF neurons expressing natriuretic peptide receptor 3 (Npr3) outside the behavioral task reproduced the low-dimensional dynamics of natural cue-evoked bursting, preserving the temporal envelope and cell-to-cell amplitude relationships across stimulation patterns. Critically, unilateral activation of non-overlapping Npr3 populations in either hemisphere recruited the full bilateral ensemble into the same native trajectory, showing that restricted inputs can nevertheless recruit the full ensemble. Matched ipsilateral and contralateral onset latencies exceeding 25 ms implicated recurrent interactions beyond the BF. Despite the conserved structure, population burst amplitude varied with stimulation strength and summed approximately linearly across concurrent unilateral inputs. This linear summation extended to natural sensory input during behavior: adding Npr3 stimulation to auditory cues increased cue-evoked burst amplitude, improved detection, and reduced subsequent reward responses, consistent with stronger reward prediction producing smaller reward prediction errors. Thus, a low-dimensional dynamical attractor preserves a shared BF burst structure across diverse inputs while integrating concurrent inputs into a scalar burst amplitude that encodes reward prediction.

Phasic reward prediction signals in ventral tegmental area (VTA) dopamine (DA) neurons and basal forebrain (BF) noncholinergic bursting neurons are remarkably similar despite the distinct circuit organizations and functions of the two systems. Whether this similarity reflects independent computation within each system, inheritance from common upstream circuitry, or causal propagation from one system to the other remains unknown. Here we show that reward prediction signals are tightly coordinated across these two systems, with VTA DA activity causally driving BF bursting neurons on a millisecond timescale. Simultaneous recordings from freely moving rats revealed strong coupling between putative DA neurons and BF bursting neurons on individual trials, both in response amplitude and spike timing. DA activity consistently preceded BF bursting neuron activity by approximately 10 ms across well-trained behavior, learning, and even outside overt task events. In DAT-Cre mice, optogenetic activation of VTA DA neurons rapidly recruited BF bursting neurons at latencies matching those observed during natural DA-BF coupling, while activation of their BF terminals produced responses at still shorter latencies. Conversely, inhibition of DA neurons or their BF terminals suppressed both cue-evoked and, notably, ongoing activity in a substantial subset of BF bursting neurons. Fast DA-evoked BF recruitment persisted after D1/D2 receptor blockade, and anatomical tracing identified glutamatergic DA-to-BF synapses, supporting glutamate co-release as a candidate substrate. Together, these findings reveal a previously unrecognized causal link that enables rapid propagation of reward prediction signals between functionally distinct subcortical systems.

Animals learn to select actions that maximize reward in complex environments. A central challenge in this process is credit routing: behavioral outcomes must be routed back to the neural populations that contributed to generating the relevant actions. One possible route is through thalamocortical pathways conveying learning signals from the basal ganglia and cerebellum, where plasticity is shaped by reward-prediction errors. However, how these pathways route such signals back to specific cortical circuits remains unclear. To address this question, we combined anatomical circuit mapping in mice with closed-loop circuit modeling. Anterograde transsynaptic tracing revealed that thalamic neurons receiving input from the substantia nigra pars reticulata or the lateral cerebellar nucleus were largely segregated, with partial overlap in the ventromedial thalamus. Nevertheless, their cortical axons shared a pronounced enrichment in layer 1a, most prominently in secondary motor cortex. In a brain- wide synergistic loop-circuit model, learning was enhanced when both pathways delivered additive and gain-modulating feedback aligned with the cortical action readout. Local activity- based calibration established this alignment, whereas dendritic-spike- and reward-prediction- error-dependent thalamocortical plasticity further refined behavioral output. Module-specific feedback was essential for a brain-machine interface task requiring the simultaneous conditioning of multiple target neurons. Together, these findings define a credit-routing architecture in which laminar convergence, plastic alignment, and modular feedback link parallel subcortical learning systems to cortical action representations.

Successful navigation requires the integration of self-motion signals with spatial representations, yet the circuits that provide velocity information necessary for stable grid cell firing remain incompletely understood. Here, we identify medial septal (MS) glutamatergic neurons as a critical component of this pathway. Using cell-type-specific optogenetic silencing combined with in vivo electrophysiology in freely moving mice, we show that MS glutamatergic neurons support the fidelity of firing rate-based speed coding in the medial entorhinal cortex (MEC). Silencing this population reduced grid cell spatial periodicity and stability and produced distortions of grid firing fields. Silencing also reduced theta phase-locking strength in strongly theta-modulated MEC neurons. Model simulations of a hybrid oscillatory interference-continuous attractor network with reductions in speed-signal fidelity disrupted grid periodicity and reproduced features of the spatial distortions observed in vivo. Together, these findings identify a septo-entorhinal circuit that coordinates self-motion information with the spatial and temporal organization of MEC activity to maintain a stable neural representation of space.

Neuronal pathology plays a prominent role in the cognitive and neuropathic symptoms of multiple sclerosis; however, little is known about these pathological signals to neurons in the brain and their consequences on cognition. Here, using a transgenic reporter of IFN signaling with spatial RNA-sequencing, we determined myelin-specific autoimmune T cells enter the brain parenchyma through the interface between the fimbria and third ventricle. Here, astrocytes and microglia express high levels of MHC-II-associated transcripts. Using this reporter, we also showed microglia within the brain are not direct responders of IFNs, but express transcripts associated with synaptic pruning, potentially contributing to pathological remodeling of neuronal circuits. Moreover, both type-I and type-II IFNs are required for T cells to infiltrate the brain and initiate an IFN response associated with synaptic and metabolic programs in neurons and astrocytes in the cortex and hippocampus. In summary, this study elucidates multiple steps of the pathogenesis of EAE pathology within the brain independent of the spinal cord.

Learning about potential threats is indispensable for survival, and its deficits underlie numerous neuropsychiatric disorders. Human brain imaging suggests the anterior insular cortex (AIC) as a key neural correlate of threat learning, yet its mechanistic role remains unclear. Leveraging multimodal systems neuroscience approaches, we demonstrate that AIC excitatory neurons, specifically layer 5 pyramidal neurons projecting to the basolateral amygdala, are essential for higher-order threat learning/memory. They are specifically required when a predictor is temporally separated from an aversive event or when the predictive information is embedded within a diffuse context. In contrast, the AIC is dispensable for elementary, temporally contiguous threat learning. Surprisingly, while the AIC promotes memory formation, it also constrains the level of fear expression during retrieval in a context-dependent manner. This bidirectional control is mediated by the recruitment of distinct AIC neuronal ensembles. Together, we identify a novel circuit mechanism for bidirectional regulation of higher-order threat learning/memory.

To understand neural computations, one would like to relate circuit dynamics to neuronal intrinsic excitability and connectivity. We developed tools for large-scale all-optical electrophysiology and applied them to approximately 300 layer 1 interneurons per mouse in vivo. Across two mice, targeted stimulation and voltage imaging identified 1,469 inhibitory and 5,752 directed electrical connections among 14,312 tested pairs. Presynaptic identity was more strongly associated with inhibitory post-synaptic potential (IPSP) amplitude, whereas postsynaptic identity was more strongly associated with IPSP decay time. Gap junctional and synaptic interactions had distinct spatiotemporal kernels, with corresponding signatures in spontaneous activity. Excitability measured under optogenetic stimulus predicted spiking responses to subthreshold voltage, depolarization rate, and spike history during spontaneous firing. Genetic subtypes also showed different behavioral-state-dependent voltage-to-spike relationships despite similar population-average subthreshold voltages. These measurements link cellular properties and pairwise interactions to spontaneous circuit dynamics, establishing an approach for functional circuit mapping in vivo.

The dorsomedial striatum guides learning and adaptive decision-making through excitatory synaptic control of its spiny projection neuron outputs. However, the contributions of local inhibitory microcircuitry remain poorly understood. Here, we identify an interneuron circuit in the dorsomedial striatum that links outcome processing to adaptive action selection. During probabilistic push-pull reversal learning, interneurons represented immediate outcomes: somatostatin interneurons were recruited on unrewarded trials and unexpected rewarded trials, whereas tyrosine hydroxylase interneurons were suppressed on unrewarded trials and recruited on rewarded trials. In vivo recruitment of tyrosine hydroxylase interneurons suppressed somatostatin interneuron activity and increased activity in both direct- and indirect-pathway striatal projection neurons, revealing a polysynaptic disinhibitory microcircuit. Transient inhibition of somatostatin interneurons in this pull-tuned region produced a sustained increase in aberrant pull choices and occupancy of a suboptimal pull-preferring behavioral state, whereas inhibition of tyrosine hydroxylase interneurons produced a sustained impairment of pull reinforcement. Longer-term policy changes following somatostatin interneuron inhibition coincided with postsynaptic potentiation of excitatory synapses onto striatal projection neurons, suggesting a potential substrate for the persistence of altered behavioral policies. Together, these findings identify a disinhibitory striatal circuit motif gating reinforcement which transforms individual trial outcomes into temporally broader policy.

Sequences of stimuli can be maintained in working memory by encoding each item, together with its position, as a distinct pattern of neural population activity. Competing hypotheses about the underlying dynamics have been proposed: patterns of activity might be supported either by stable attractor states that persist over time, or by transient sequential dynamics. To investigate the dynamics underlying sequence memory, we developed a framework for fitting low-rank recurrent neural networks (RNNs) to multi-session recordings of spiking activity, and applied it to 2,488 units recorded across 30 sessions in two macaques. Our data-constrained RNNs generated population spiking activity which closely resembled neural data and recovered single-unit tuning properties across sessions. Consistent with prior work, fitted networks represented items presented at different sequence positions in separate low-dimensional subspaces in the neural activity, alongside a dominant temporal component. The RNNs enabled direct investigation of the underlying dynamics: By isolating the temporal and stimulus subspaces, we found a manifold of fixed points emerged during the memory retention period. Targeted perturbation analyses confirmed that activity within this manifold is consistent with attractor-like dynamics, and predict that brief optogenetic perturbations delivered along this attractive manifold can selectively induce persistent shifts in neural activity and decoded behavior. Together, our results indicate that sequence working memory is organized by position-specific coding subspaces, within which attractor states maintain persistent representations, while activity evolves along time-varying dimensions.

Dopamine in the nucleus accumbens Core powerfully influences motivation and learning, yet the information encoded by Core dopamine remains unresolved. Competing theories typically assume a spatially uniform signal, and standard recording methods average dopamine over substantial tissue volumes. Using two-photon imaging through microprisms in behaving mice, we reveal a pronounced spatial organization of Core dopamine dynamics. Posterior Core showed phasic dopamine increases to both rewards and aversive stimuli, with transfer to a predictive cue after learning. Anterior Core showed phasic decreases to aversive stimuli, weaker increases to reward, and higher tonic dopamine that was further amplified by cocaine. Rather than uniform or highly fragmented signals, the Core contains a smoothly graded mixture between two distinct sets of dopamine dynamics, neither matching canonical reward prediction errors.

Astrocytic processes closely associate with excitatory synapses and express AMPARs, positioning them to sense the same glutamate signals as neurons. Whether a mismatch between neuronal and astrocytic glutamate sensing affects synaptic physiology is unknown. Using AstroHTP, an extension of Drug Acutely Restricted by Tethering (DART) for cell-type-specific pharmacology, we found that acute astrocytic AMPAR inhibition spared neuronal AMPARs and basal transmission yet converted ongoing glutamatergic activity into a remarkably potent form of long-term depression (LTD) in mouse striatal projection neurons (SPNs). Unlike canonical presynaptic LTD, this form of LTD manifests postsynaptically, requires non-ionotropic NMDA receptor signaling, and arises when glutamate transmission occurs without a matching astrocytic signal. Astrocytic glutamate sensing through AMPARs, therefore, acts as a parity signal that verifies transmission and stabilizes active synapses.

Animals can solve the same problem using distinct strategies, ranging from memorization of specific sensory details to inference based on abstract task structure. Although the hippocampus has been implicated in cognitive map formation, how the structural information emerges in the hippocampus through learning and whether it is associated with behavioral strategies remain poorly understood. Here, we developed a structured cue-guided navigation task in virtual reality and performed longitudinal two-photon calcium imaging of hippocampal CA1 neurons throughout learning in mice. Although the task did not explicitly require learning of latent structure, most animals progressively transformed hippocampal map from sensory-oriented representations into cue-invariant representations of reward sequences. Strikingly, the emergence of such abstract hippocampal maps predicted whether individual animals adopted structure-based behavioral strategies when sensory information became partially unavailable. At the population level, reward-sequence representations formed a latent-state map capable for predicting future state transitions, generating a parsimonious Markov graph. Computational modeling further identified next-state prediction and representational regularization as two computational principles essential for generating similar latent-state structures across distinct architectures. Together, our findings suggest that the hippocampus spontaneously extracts abstract task structure through learning and that the geometry of hippocampal representations predicts problem-solving strategies.

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease with ~90% of cases being sporadic. Although genome-wide association studies have identified numerous genetic risk loci, these variants account for only a portion of ALS heritability, suggesting that additional genetic and regulatory mechanisms contribute to disease risk. Furthermore, the mechanisms linking common variants to disease pathogenesis remain poorly understood. Adenosine-to-inosine (A-to-I) RNA editing is dysregulated in ALS, yet systematic identification of disease-relevant editing events and their functional consequences has been limited. We performed comprehensive RNA editing analysis across eight central nervous system regions from a Target ALS cohort, identifying 752 differentially edited sites in 304 genes. Global RNA editing was significantly reduced in ALS tissues, particularly in spinal cord and choroid plexus, with altered editing in genes involved in immune signaling, stress response and RNA processing. Through genetic colocalization analysis integrating editing quantitative trait loci (edQTLs) with ALS GWAS data, we identified common variants at the DEGS1 locus associated with both increased 3' UTR editing and elevated ALS risk (posterior probability >0.88). DEGS1 encodes dihydroceramide desaturase 1, a key enzyme converting dihydroceramide to ceramide, the precursor to all sphingolipid species. Using CRISPR-generated DEGS1 knockout iPSCs and lipidomic profiling, we demonstrated that DEGS1 is essential for ceramide synthesis and downstream sphingolipid production in astrocytes. Functional validation revealed that ALS-associated editing of the DEGS1 3' UTR significantly increased DEGS1 protein expression. Critically, astrocytes expressing the fully edited DEGS1 3' UTR increased ceramide levels and reduced cell viability compared to those expressing the unedited construct. These findings establish a novel regulatory link between RNA editing, sphingolipid metabolism and ALS pathogenesis. Our results demonstrate that common genetic variants can influence disease risk through post-transcriptional mechanisms that alter lipid homeostasis in astrocytes, a cell type increasingly recognized as central to ALS progression.

The electrical output of a neuron is generated and conducted along its axon, yet the axon has remained largely inaccessible to direct measurement of membrane potential in intact tissue. Genetically encoded voltage indicators have enabled optical recording of membrane potential dynamics from neuronal cell bodies and dendrites in vivo, but direct voltage imaging of identified axons in the living animal has remained largely unexplored. This gap is especially consequential for nociceptive neurons, whose signals are generated and propagated at peripheral terminals far from the soma. Here, we expressed genetically encoded voltage indicators in trigeminal neurons and, using high-speed widefield imaging with holographic targeted illumination, recorded membrane-voltage dynamics directly from corneal nociceptive terminal arbors and their parent axons in vivo, in intact tissue. Individual terminals responded to capsaicin and menthol with depolarization accompanied by trains of spiking activity. Unexpectedly, a substantial fraction generated ongoing action-potential-like firing in the absence of an applied stimulus. This firing was sodium-channel-dependent, synchronized across sister terminal branches, and shared its dynamics with the parent axon. Ongoing terminal firing persisted during lidocaine delivery to the trigeminal ganglion, supporting a peripheral origin distal to the ganglion, and was abolished by inhibition of NaV1.8 or HCN/Ih conductances. Thus, direct voltage imaging reveals ongoing and spatially coordinated electrical activity within intact peripheral sensory arbors. More broadly, this approach provides access to the generation and integration of neuronal output within fine axonal compartments in vivo.

The ability to recover from consecutive failures is critical for adaptive behavior, yet the underlying neural mechanisms remain unclear. In vivo screening of neuromodulatory inputs onto frontal-sensory cognitive control circuit identified a frontal cholinergic mechanism that counteracts lapses in attentional engagement following repeated errors. Cholinergic input from the basal forebrain to the frontal cortex was necessary and sufficient to restore performance after errors, specifically after repeated-but not after a single-failure. This effect was mediated by 7-nicotinic acetylcholine receptors in frontal-sensory cortical neurons, which sustained cortical circuit activity and frontal-sensory cortical oscillations during attentional engagement. Targeted optogenetic activation of these frontal-sensory neurons disrupted cycles of attentional performance failure and reinstated behavioral performance in a mouse model of neuropsychiatric syndrome. These findings reveal a circuit-level mechanism of cholinergic control that confers cognitive resilience under conditions of persistent failure.

The transition from single-neuron excitability to organized population firing is a fundamental feature of neural circuits, yet the mechanisms that govern this transition remain poorly understood. Magnocellular vasopressin (VP) and oxytocin neurons are intermingled within the hypothalamus, yet display strikingly different population dynamics, with VP output sustained by a distributed, largely asynchronous activity. Here, we identify a local mechanism by which somatodendritically released VP acts as a diffusible signal that converts an intrinsic excitability mechanism into spatially and temporally organized feedback across the VP population. Activity-dependent somatodendritic VP release recruited a rapid autocrine feedback that transiently potentiated the slow afterhyperpolarization (sAHP) through V1aR signaling which in turn strengthened spike-frequency adaptation and restrained firing. The same signal also acted in a diffusible, paracrine manner on neighboring VP neurons, producing distance- and time-dependent sAHP modulation, with rapid potentiation at short range and delayed inhibition at greater distances. Focal VP uncaging further revealed preferential engagement of somatic over dendritic signaling. Together, these findings show how a diffusible neuropeptide can convert an intrinsic excitability mechanism into spatially structured intercellular feedback, providing a candidate mechanism for sustaining distributed, asynchronous VP population output during prolonged homeostatic demand.

Oxytocin (OXT) can exert diverse effects across social, affective and motivational domains, yet its impact on brain function at the systems level remains unclear. Here, we combine chemogenetics with fMRI and electrophysiology in mice to investigate how endogenous OXT reshapes intrinsic large-scale brain network dynamics. Using mice constitutively expressing hM3Dq DREADD receptors in OXT-releasing hypothalamic neurons, we show that chemogenetic stimulation of OXT neurons elicits widespread increases in cerebral blood volume, with prominent effects in fronto-striatal regions and dorsal hippocampal areas. This broad functional response is accompanied by a network-specific reconfiguration of brain-wide functional connectivity, characterized by increased coupling between key components of the rodent social brain, and negative functional coupling between hippocampal and fronto-cortical regions. In vivo electrophysiology under matched conditions shows concordant changes in slow (<1 Hz) and delta-band coherence across frontalthalamic and fronto-hippocampal circuits, revealing a plausible neural correlate of the observed connectivity effects. Notably, these changes are accompanied by a marked reconfiguration of the temporal architecture of fMRI network activity, with increased occupancy of co-activation states engaging saliencerelated regions and a concomitant suppression of sensory-dominated states. Together, these findings identify dynamic network-state reconfiguration as a systems-level mechanism through which endogenous OXT can tune functional communication across social-affective circuits.

Animals continuously transform sensory inputs into representations that guide behavior. Although the amygdala is a brain structure critical for learning and choice, how natural sensory experience is represented within this circuit remains unknown. Here, we identify a representation of the visual environment in the human basolateral amygdala (BLA) that contains information relevant for learning and choice. Using image-computable encoding models trained on responses to naturalistic videos, we recover a continuous, low-dimensional representation of visual experience in the BLA. The representation generalizes across individuals and independent datasets and prospectively predicts BLA responses to novel visual stimuli generated through closed-loop image synthesis. Signals derived from the representation also informed reinforcement-learning models of human choice. Together, these findings show that the human BLA organizes complex visual experience into a structured representation that can support learning and flexible behavioral responses.

Internal states such as hunger dynamically reshape activity across circuits to support resource seeking. Neuromodulation provides a means of controlling such physiological properties of neurons, but how this flexibility regulates memory networks remains unclear. Here, we describe how hunger reconfigures a dopaminergic food-reward circuit between two physiological modes that support memory formation, in Drosophila. Starvation suppresses baseline dopaminergic activity through peptidergic signalling, enabling reward-evoked, large-amplitude dopamine neuron spikes to reinforce learning. This spiking mode can be engaged by sugar consumption and persists beyond feeding, reflecting the fly's satiety state. Persistent dopaminergic large-amplitude spiking reinforces learning and transitions the memory network into a mode that prioritizes consolidation of recently acquired memories. The transition from hunger to satiation is also reflected in the activity of postsynaptic output neurons that shift from a tonic, decorrelated mode that is responsive to dopamine into a bursting, correlated mode in which further dopaminergic modulation is occluded. Therefore, nutrient deprivation reconfigures the memory circuit into a learning competent dopamine-receptive mode, which persistent reinforcing dopamine then switches into a satiated mode driving memory consolidation. Together these processes mechanistically intertwine state-dependent reward-signalling with subsequent memory stabilization through transitions in physiological mode.