bioRxiv Subject Collection: Neuroscience's Journal

Injury-Induced Remodeling of Junctional Actin Bands in the Vestibular Maculae of Mice and Chicks: Implications for Sensory Regeneration
The vestibular organs of birds are capable of regenerating sensory hair cells after ototoxic injury, but the regenerative ability of the mammalian vestibular organs is much more limited. The factors that inhibit regeneration in the mammalian inner ear are not known, but it has been proposed that the structure of filamentous actin cables at cell-cell junctions within the sensory epithelium may be an important regulatory influence. Junctional actin cables in the chick utricle are relatively thin, while those in mouse utricle are much thicker. These differences result in differing mechanical properties of the avian vs. mammalian inner ear, which may affect the potential for regenerative proliferation. The present study characterized injury-evoked changes in junctional actin cables in the utricles of mice and chicks. We found that the thickness of junctional cables in the chick utricle was not affected by ototoxic injury, but that injury to the mouse utricle led to the formation of many new junctional actin bands whose thickness was comparable to those in the chick utricle. Thicker actin bands persisted after injury, but were not necessarily associated with cellular junctions. In addition, the relative extent of supporting cell expansion in the injured chick utricle was larger than that in the mouse utricle, which may affect activation of Hippo/YAP signaling in both species. Together, these data point to important differences in actin cable plasticity in the avian vs. mammalian utricle that may partially account for their differing regenerative abilities.

In search for the invisible: motor inhibition in monkey premotor cortex and its RNN replicas
Controlling actions in dynamically changing environments requires flexible and efficient motor control. A fundamental challenge in neuroscience is to uncover how cortical circuits generate, adjust, and sometimes suppress planned movements. To address this, we combined recordings from dorsal premotor cortex (PMd) of macaque monkeys performing a stop-signal task, with a recurrent neural network (RNN) model inferred directly from multi-unit activity. This data-driven "digital twin" reproduced the cortical population dynamics underlying motor planning and inhibition, revealing how internal network states shape behavior, and generating synthetic neural trajectories for unseen conditions. RNN internal state explained reaction time fluctuations across trials, reflecting stochastic components of motor readiness and endogenous variability of PMd activity. The same pre-Go latent state also constrained movement inhibition modulating the network's response to Stop signals by reshaping the attractor dynamics of a null-potent subspace. These results establish a mechanistic link between latent cortical dynamics and flexible behavioral control, demonstrating how autonomous RNN inference can uncover circuit-level computations.

Serotonin modulates nucleus accumbens circuits to suppress aggression
Serotonin (5-hydroxytryptamine; 5HT) has long been considered anti-aggressive, but the mechanisms by which 5HT regulates downstream circuits to control aggression remain unclear. Combining fiber photometry, optogenetics, and miniaturized microscope recordings in double-transgenic mice, we find that 5HT levels ramp up in the nucleus accumbens during aggression, inhibiting a subset of D1 medium spiny neurons to suppress attacks. Our results reveal a novel 5HT-mediated neuromodulatory mechanism for limiting aggressive behavior.

Coordinated multilaminar dynamics underlie multiplexed computation in motor cortex
Functional multiplexing is a signature of higher-order brain regions. Beyond single-unit mixed selectivity, the six-layered cortical microcircuit has been proposed as an optimal substrate for such parallel computations. Recurrent connections within and across layers allow cortical columns to retain incoming information for some time and integrate it into stable latent representations, thereby acting as functional units. The key question is whether the multiplexing capacity arises from specialized processing within individual layers or the collective coordination of multi-layer activity patterns. Here, we analyzed laminar recordings from the motor cortex of macaque monkeys performing a complex delayed match-to-sample task. We identified laminarly distributed, behaviorally specific subspaces that captured the encoding of distinct task-related variables. These subspaces, spanning the entire column and expressed as coordinated activity patterns, were functionally reused to encode the same variable over time and flexibly recycled to encode new ones. Subtle variations of laminar weights gave rise to multiple coexistent laminar coding subspaces, enabling multiplexing at the columnar level. Task-related information propagated across layers in temporally organized trajectories that transiently localized in superficial or deep layers at distinct trial epochs. These organized laminar trajectories were consistently observed across recording sites, but exhibited site-specific propagation patterns. The activity of the population on the other hand lacked structured dynamics. Thus, laminar trajectories of information emerged atop a background of spatially and temporally unspecific activity-fluctuations.

Language usage modulates the neural mechanisms ofselective attention in bilinguals
The use of multiple languages modulates the neural mechanisms of selective attention, but it is unclear whether these adaptations require continuous engagement and active second language usage. Here we examined whether language usage shapes selective attention in bilingualism, and what neural processes might this engage. 48 highly proficient English-French bilinguals listened to naturalistic speech streams in their first or second language, paired with either linguistic and non-linguistic interference. Participants were matched on their L2 proficiency, but were either Active, Moderate, or Inactive users of their second language. The results revealed usage-related modulation of oscillatory activity in the alpha band, with more efficient inhibitory control in Active and Moderate users leading to behavioural resilience to interference. In contrast, usage did not affect lower-level perceptual tracking of speech, as captured by mTRF decoding of speech envelopes across frequency bands. Taken together, our findings show that resilience to interference during language processing is not dependent on the perceptual speech tracking, but rather on the capacity to recruit higher-level attentional control mechanisms, a process that is dynamically shaped by bilinguals L2 usage.

Decoding the neural stages from action and object recognition to mentalizing
Higher-level action interpretation, such as inferring underlying intentions and predicting future actions, requires the integration of conceptual action information (e.g. "opening") with semantic knowledge about persons and objects (e.g. "my friend Anna", "pizza box"). However, how the neural systems for action and object recognition and memory interact with each other to form the basis for inferring higher-level mental states remains unclear. Here we use fMRI-based crossmodal multiple regression representational similarity analysis in human female and male participants to elucidate the processing stages from basic action and object recognition to mentalizing. We show that inferring intentions from observed actions or written sentences involves a modality-general network of lateral and medial frontoparietal and temporal brain regions associated with conceptual action and object representation and mentalizing. The representational profiles in these regions are explained by models capturing different types of conceptual information, revealing distinct but partially overlapping networks for action, object, and mental state representation. There was no strict separation of networks for action, object, and mental state representations, arguing against a sequential bottom-up hierarchy from action and object understanding pathways to the mentalizing network. Rather, left-hemispheric regions, specifically ventrolateral prefrontal, inferior parietal and anterior lateral occipitotemporal cortex, showed strong representational overlap, pointing towards a core network for making meaning of action-object structures at a conceptual level. We argue that this core network represents a distributional semantic hub between classic networks for action and object understanding and the mentalizing network.

Significance StatementHow does the human brain integrate information from actions, e.g., "open a pizza box", to understand the actions underlying intentions? To do so, the brain needs to combine information from different neural networks--for action and object recognition--and pass them to the mentalizing network for inferring intentions, such as "satisfying hunger". We characterize the interplay of networks using fMRI-based crossmodal multivariate analyses and find that a left-lateralized core network in inferior frontal and parietal cortex and lateral occipitotemporal cortex represents all critical ingredients--conceptual action and object information as well as higher-level mental state representation simultaneously in an overlapping manner. This suggests that this core network is essential for semantic interpretation and functions as bridge between recognition pathways and the mentalizing system.

A cholinergic mechanism orchestrating task-dependent computation across the cortex
In an ever-changing environment, animals often need to switch between performing different tasks involving distinct sets of cognitive processes. Many such tasks involve neural activity distributed across the cortex, with dynamics that depend on both task demands and behavioral strategy1-17. A fundamental but unanswered question is what circuit mechanisms orchestrate these task-dependent dynamics. Here, we hypothesized that acetylcholine release in the cortex plays a key role. The cortexs only long-range source of this neuromodulator is the basal forebrain cholinergic system, which targets the entire cortical sheet and can individually modulate single regions on sub-second timescales18-26. To test our hypothesis, we first imaged cholinergic axons innervating the cortex while mice switched frequently between two navigational decision-making tasks in virtual reality (VR), only one of which required gradual accumulation of sensory evidence. We found that cholinergic input to the cortex is spatiotemporally heterogeneous and multiplexes sensory, motor, arousal, and cognitive signals in a task- and strategy-dependent fashion, with overall higher activity during evidence accumulation. Crucially, beyond contextual variables, cholinergic activity directly tracked task computations themselves, encoding an evidence-dependent decision variable only in the accumulation task. To test if acetylcholine release is causal to the performance of each task, we optogenetically silenced cholinergic terminals in the cortex while simultaneously imaging excitatory cortical activity. We found that this input is selectively required for evidence accumulation, and for large-scale cortical coding of evidence and choice during the accumulation task. Thus, we have identified a new cholinergic mechanism that orchestrates cortex-wide activity in a task-dependent manner and serves as a key node in the distributed brain network underlying the accumulation of sensory evidence.

Inferential planning in the frontal cortex
How the brain plans and maintains sequences of future actions remains a central question in systems neuroscience. Recent studies in the frontal cortex have revealed that multiple elements of a sequence are represented simultaneously in separable neural subspaces, challenging classical serial models of sequential planning. Here, we show that these representations emerge naturally under inferential planning in which sequential actions are inferred from sensory evidence and goals. Using a hierarchical generative model, we reproduce key neural phenomena observed in primate frontal cortex, including the simultaneous activation of multiple plan elements, the emergence of (almost) orthogonal `memory' subspaces, and their reuse across forward and backward sequence tasks. Our approach provides a mechanistic account of how probabilistic inference over control states gives rise to distributed and dynamic neural representations of plans. This framework not only unifies previously disparate findings on planning, working memory, and motor preparation, but also generates novel, testable predictions about the dynamics of active inference, the role of sensory subspaces, and the impact of uncertainty on sequence processing.

Dynamic brain states during encoding and their post-encoding reinstatement predicts episodic memory in children
Episodic memory, the ability to rapidly learn and explicitly remember past events and experiences, plays a critical role in childrens academic learning and knowledge acquisition. The formation of lasting memories relies on the brains ability to dynamically organize its activity. How these neural configurations unfold moment-by-moment across encoding and offline phases remains poorly understood. To probe these dynamics, we applied a novel Bayesian Switching Dynamic Systems approach, a hidden Markov model with automatic state detection, to fMRI data from children performing scene encoding followed by an offline post-encoding rest. We identified four distinct brain states during encoding with unique activation modes between visual, medial temporal lobe, and frontoparietal and default mode network nodes. An "active-encoding" state with integrated visual-hippocampal and frontoparietal activity dominated encoding and predicted individual memory performance, while an inactive state negatively predicted performance. State transition dynamics revealed that flexible shifts into the active-encoding state enhanced memory formation, whereas transitions toward inactive states impaired it, demonstrating that memory success depends on dynamic neural flexibility. Critically, encoding states spontaneously reemerged during post-encoding rest. A "default-mode" state characterized by enhanced default mode network activity showed sustained maintenance during rest and robustly predicted memory outcomes, an effect specific to post-encoding, not pre-encoding rest. These findings establish that episodic memory emerges from coordinated brain state sequences bridging online encoding with offline consolidation, providing a computational framework for how moment-to-moment neural dynamics support memory formation in children. This work has broad implications for optimizing educational interventions and understanding developmental disorders affecting learning and memory.

A flexible coding scheme underlying working memory generalization in human parietal and frontal cortices
Humans can rapidly extract abstract, common knowledge from distinct experiences, enabling generalization of this knowledge across tasks within working memory (WM) to guide behavior. Although task-specific representations of WM have been observed in a distributed network encompassing sensory, parietal, and frontal cortices, whether these regions construct task-generalizable representations and whether such representations can be dynamically shaped by task demands remain unclear. Across two functional MRI experiments, we investigated WM generalization across distinct WM tasks (location and object tasks) based on shared underlying task structures. Trial-specific task demands varied between passive maintenance and active, rule-guided manipulation of mnemonic stimuli. The posterior parietal cortex (PPC) demonstrated task-generalizable stimulus representations during both maintenance and manipulation. In contrast, the lateral (LFC) and medial (MFC) frontal cortices supported manipulation-and maintenance-based generalization, respectively. Critically, manipulation-based generalization in the PPC and LFC emerged even when participants did not explicitly learn the mapping between task spaces, indicating that active exploration of task structure can spontaneously facilitate generalization across tasks. Together, these findings reveal that flexible generalization for goal-directed behavior is achieved via a distributed WM network, with distinct regions serving active versus passive task demands.

Ventral Hippocampal Temporoammonic and Schaffer Collateral Pathways Differentially Control Fear- and Anxiety-Related Behaviors
The ventral hippocampus plays a crucial role in regulating anxiety- and fear-related behaviors. Previously, we demonstrated that diazepam reduces anxiety-like behavior by inhibiting the dentate gyrus and CA3 principal neurons via 2-GABAARs, while inhibition of CA1 pyramidal neurons is necessary to suppress fear-related responses. This study investigated the role of inputs from ventral CA3 (vCA3) and entorhinal cortex to ventral CA1 (vCA1) in anxiety- and fear-like behavior using bidirectional optogenetic modulations. Adult C57BL/6J male and female mice were subjected to bilateral stereotaxic injection of a viral vector expressing channelrhodopsin or halorhodopsin into vCA3 or into layers II-III of lateral entorhinal cortex, followed by bilateral implantation of fiberoptic ferrules into vCA1. After four weeks of recovery, mice were assessed for anxiety-like behavior in the novel open field, elevated plus maze, and Vogel conflict tests, and by contextual and trace fear conditioning for fear. The behavior of the mice was recorded under laser ON and OFF conditions in all experiments. The activation of vCA3 to vCA1 projections (i.e., Schaffer collateral pathway) increased anxiety- and fear-related behaviors, whereas inhibition reduced such behaviors. In contrast, optogenetic activation or inhibition of EC to vCA1 projections (i.e., temporoammonic pathway) had no effect on anxiety-related behavior but positively or negatively modulated fear-related behavior, respectively. These results suggest that while fear-related behavior is modulated by both inputs to vCA1, modulation of anxiety-related behavior is input-specific for the vCA3 to vCA1 projection. In summary, this study offers mechanistic insights into the complex organization of hippocampal circuitry underlying fear and anxiety.

GRAPHICAL ABSTRACT

O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=183 SRC="FIGDIR/small/690696v1_ufig1.gif" ALT="Figure 1">
View larger version (42K):
org.highwire.dtl.DTLVardef@1041871org.highwire.dtl.DTLVardef@1e0bde8org.highwire.dtl.DTLVardef@1d3dab1org.highwire.dtl.DTLVardef@1bac990_HPS_FORMAT_FIGEXP M_FIG C_FIG

Early Arousal Signals Drive Reward Learning and Subsequent Choice Behaviour
Learning to reinforce rewarding actions and avoid repeated mistakes is crucial for survival in dynamic environments. Yet, it remains unclear how distinct neural signals coordinate to implement reward-based decision-making and behavioural adjustment. We obtained simultaneous electroencephalography (EEG) and pupillometry during a probabilistic reversal learning task. Leveraging single-trial EEG, we first replicate the presence of two feedback-locked neural representations; an early signal previously linked to alertness and switching behaviours following negative feedback and a late signal associated with value updating and reward learning. Using single-trial pupillometry, we then show that differences in feedback-evoked pupil responses between positive and negative feedback are driven primarily by negative feedback encoding. Jointly examining these EEG and pupillometry signatures, we show that following negative feedback, increased trial-by-trial coupling between the pupil response and the early, but not the late, EEG signal is linked to increased uncertainty and exploration tendency as well as reduced accuracy and evidence accumulation on the next trial. Consistent with previous research implicating the locus-coeruleus-noradrenaline system in uncertainty signalling and network resets, we propose that when internal estimates of contextual uncertainty are high following negative feedback, an early signal, likely regulated by locus coeruleus activity, implements a network reset in reward learning structures of a later learning signal. This interruption may simultaneously increase the neural gain related to the processing of novel information and decrease the influence of existing representations in reward learning structures, in turn improving performance by creating new, more accurate internal representations of the external world.

Significance StatementThe current study jointly examines EEG and pupillometry signatures associated with reversal-learning during reward-based learning. It suggests that when internal estimates of contextual uncertainty are high following negative feedback, an early neural signal, likely regulated by locus coeruleus activity, implements a network reset in reward learning structures of a later learning signal. This interruption may simultaneously increase the neural gain related to the processing of novel information and decrease the influence of existing representations in reward learning structures, in turn improving performance by creating new, more accurate internal representations of the external world.

Enhanced pitch perception in early blind individuals and musicians is due to reduced internal noise
Although it is commonly assumed that early blind individuals have enhanced pitch perception, the literature to date has been contradictory. We measured performance in two tasks known to recruit early vs. late stages of auditory frequency tuning in early blind and sighted individuals with matched levels of musical training. The two groups showed equivalent performance on a tone detection in a notched-noise task that assesses peripheral and subcortical frequency selectivity. In contrast, early blind individuals showed enhanced performance in a pitch discrimination task thought to be mediated by cortical mechanisms. Computational modeling revealed that this enhancement was best explained by reduced internal noise rather than narrower frequency tuning, with both blindness and musical training predicting a reduction in internal noise. These findings identify internal noise as a key factor in experience-driven auditory plasticity as a result of both musical experience and early blindness.

Oscillatory brain activity reflects semantic and phonological activation during sentence planning.
Verbal short-term memory includes resources for maintaining semantic and phonological information. These resources are complementary and often activated simultaneously, making their anatomical bases difficult to determine. We mapped the neural representation of semantic and phonological short-term memory by recording magnetoencephalography (MEG) data while participants rehearsed short formulaic 5-word sentences like "The mouse ate the cheese." During a memory delay period, participants exhibited bilateral temporofrontal event-related desynchronization (ERD, power decrease) in the alpha and beta bands (8-30 Hz). During the memory delay, participants also heard an auditory distractor word that could be unrelated to the sentence, semantically related to one of the words in the rehearsed sentence (e.g. "rat" or "butter"), or phonologically related to one of the words in the sentence (e.g. "mountain" or "cheap"). Relative to unrelated words, related words induced a greater degree of ERD immediately following their presentation. Effects of semantic distractors were exclusively in the temporal lobe, largely in the left middle temporal gyrus but also in bilateral medial temporal regions. Effects of phonological distractors were far more widespread in temporal, frontal, and parietal regions, and were largely left-lateralized. As no behavioural effects were observed in cued sentence repetition, it seems that auditory distractors produce short-lasting interference with a verbal memory trace that is ultimately resolved, but useful for mapping regions involved in maintaining distinct aspects of the sentence content.

AKAP1 regulates mitochondrial and synaptic homeostasis to enable neuroprotection and repair in retinal ganglion cell degeneration
Glaucoma is a leading cause of irreversible blindness, characterized by progressive retinal ganglion cell (RGC) loss and optic nerve degeneration. Mitochondrial dysfunction plays a central role in this neurodegeneration, yet effective targeted therapies remain limited. Here, we identify the mitochondrial scaffold A-kinase anchoring protein 1 (AKAP1) as a critical regulator of RGC resilience and axon regeneration. AKAP1 expression is diminished in human glaucomatous retinas and experimental glaucoma models, correlating with elevated intraocular pressure, disrupted mitochondrial dynamics, oxidative stress, and synaptic instability. Restoration of AKAP1 via adeno-associated virus serotype 2-mediated gene therapy preserves RGC survival, promotes mitochondrial fusion and cristae integrity, enhances ATP production, and mitigates oxidative and apoptotic stress in mouse models of glaucoma and optic nerve injury. Transcriptomic profiling of AKAP1 knockout retinas reveals widespread dysregulation of mitochondrial and synaptic gene networks. Mechanistically, AKAP1 stabilizes synapses by promoting mitochondrial biogenesis, modulating calcium/calmodulin-dependent kinase II and synapsin phosphorylation, maintaining synaptophysin expression, and suppressing complement component C1q expression, thereby preventing early synaptic loss in glaucomatous neurodegeneration. Moreover, restoring AKAP1 expression facilitates axonal regeneration, preserves the central visual pathway, and maintains visual function. Collectively, these findings establish AKAP1 as a master regulator of mitochondrial and synaptic homeostasis and axonal regeneration and a promising therapeutic target for vision preservation in glaucomatous neurodegeneration.

One Sentence SummaryAKAP1 protects retinal ganglion cells and preserves vision by restoring mitochondrial and synaptic health in experimental glaucoma models.

Muscle-specific DNM2 overexpression improves Charcot-Marie-Tooth disease in vivo and reveals a narrow therapeutic window in skeletal muscle
Charcot-Marie-Tooth disease (CMT) caused by dominant loss-of-function mutations in DNM2, encoding the GTPase dynamin-2, impairs motor and sensory function. However, the respective contributions of muscle and nerve pathology, and the therapeutic potential of increasing DNM2 expression, remain unresolved. We evaluated tissue-targeted and systemic approaches to increase DNM2 in a mouse model carrying the common K562E-CMT mutation. Muscle-specific DNM2 overexpression from embryogenesis in Dnm2K562E/+ mice ameliorated desmin and integrin mislocalization, membrane trafficking defects, mitochondrial abnormalities, and fibrosis in skeletal muscle, resulting in improved locomotor performance despite persistent muscle atrophy. Conversely, systemic postnatal AAV delivery of human DNM2 increased DNM2 in muscle but failed to transduce nerves, and paradoxically worsened the muscle pathology, producing centronuclear myopathy-like features. These findings reveal a primary pathogenic impact of DNM2-CMT mutation within skeletal muscle, independent of nerve involvement. Collectively, they underscore that precise DNM2 dosage is critical for neuromuscular homeostasis and reveal a narrow therapeutic window for safe and effective therapeutic intervention. This paradox, in which efforts to compensate for a loss-of-function neuropathy risk inducing a gain-of-function myopathy, highlights the need for tightly controlled modulation of DNM2 activity in future therapeutic strategies.

Ataxin-2 knockdown is neuroprotective via cell-autonomous and non-cell-autonomous mechanisms
Ataxin-2 (Atxn2), a ubiquitously expressed RNA-binding protein, has been implicated in ALS risk, and its silencing represents a promising therapeutic strategy to extend life span and ameliorate symptoms in ALS. Although neuroprotective effects have been shown in several ALS preclinical models, the molecular mechanisms underlying ataxin-2 downregulation neuroprotective effects remain poorly understood. Starting with a global proteomic profiling of the yeast and mouse TDP-43 models, we uncovered that pbp1/ataxin-2 downregulation rewires metabolism to activate alternative cellular pathways for energy production under stress. By combining proteomic profiling insights and functional studies in neural cultures, we further show that ataxin-2 downregulation adjusts metabolism in both cell-autonomous and non cell-autonomous pathways. In neurons, ataxin-2 downregulation activates glycolysis and reductive glutamine carboxylation, while astrocytes provide enhanced support by increasing cholesterol synthesis. Collectively, our data characterize novel cellular pathways to overcome TDP-43 toxicity and establish ataxin-2 as a regulator of adaptive response in brain cells under the conditions of stress.

Synaptopodin KO rat for assessing the dendritic spine apparatus and axonal cisternal organelle in synaptic plasticity, development, and behavior
The actin-binding protein synaptopodin (Synpo) regulates the cytoskeleton and intracellular Ca2+ and is important for long-term potentiation (LTP) and learning. The inconsistent onset age for LTP in mice makes their Synpo knockout (KO) a suboptimal developmental model. Hence, we generated Synpo KO rats using CRISPR-Cas9. Synpo KO rats are viable with reduced body weight and bone length after postnatal days (P)35-P45. Their basal kidney function is normal. 3D reconstruction from electron microscopy reveals the absence of the Synpo-dependent dendritic spine apparatus and cisternal organelles in the axon initial segment (AIS), which may contribute to reduced LTP in the KO rat. Inhibitory synapses in the wild-type AIS appear preferentially clustered near cisternal organelles--a pattern disrupted in the KO, where synapses appear more uniformly distributed. The consistent developmental profile of LTP in the rat makes this KO a robust model to assess Synpo function in development, synaptic plasticity, and behavior.

Graphical Abstract

O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/690444v1_ufig1.gif" ALT="Figure 1">
View larger version (47K):
org.highwire.dtl.DTLVardef@43bc29org.highwire.dtl.DTLVardef@f652f7org.highwire.dtl.DTLVardef@1c0ec2aorg.highwire.dtl.DTLVardef@d0c281_HPS_FORMAT_FIGEXP M_FIG C_FIG

Nutrient stress activates Rab5b-mediated autophagy to remodel the synaptic proteome
Synaptic proteostasis is crucial for maintaining neuronal function and plasticity, yet how synapses adapt to metabolic stress remains poorly understood. Here, we show that nutrient deprivation, particularly serum withdrawal, induces robust autophagy-dependent remodeling of the synaptic proteome, while mTORC1 inhibition has more limited effects. Nutrient stress rapidly activates autophagy both globally and at synapses, with synaptic autophagy peaking within 1-2 hours of serum withdrawal. Mechanistically, we uncover that the LC3 lipidation complex (ATG5-ATG12-ATG16L1) is recruited to synapses via Rab5b-positive endosomes in a dynein-dependent manner. Live imaging reveals enhanced Rab5b-ATG16L1 co-trafficking and increased ATG5 mobility upon serum withdrawal, supporting a model of spatiotemporally controlled autophagy precursor delivery to synaptic compartments. Functionally, nutrient deprivation acutely dampens neuronal excitability in vitro, while a two-week fasting-mimicking diet in vivo triggers synaptic proteome remodeling that overlaps with starvation-induced autophagy cargo. In contrast, restriction of mTORC1-activating amino acids fails to induce comparable synaptic changes, suggesting that synaptic autophagy is regulated by nutrient signals beyond mTORC1. Our findings define a Rab5b-mediated trafficking mechanism that couples nutrient sensing to localized synaptic degradation, providing new insight into how neurons preserve proteostasis under metabolic challenge.