Biology subjects

Isomura, Y.

Publications and source records attributed to Isomura, Y..

15 recordsLinked to original sources

bioRxiv2026

Recurrent Spiking Neural Networks for Continuous Motor Control: A Biologically Plausible Model of Motor Cortex

Motor cortex acquires skilled movements through recurrent spiking dynamics and experience-dependent synaptic plasticity, yet how such circuits learn continuous motor control using biologically local signals remains unclear. Here, we develop a recurrent spiking neural network model of motor cortex that addresses both the computational and biological requirements of motor learning. We first analytically derived the recurrent firing-rate state directly from synaptic input and trained the network by implicit differentiation, thereby avoiding backpropagation through long spike trajectories. We then introduced State-Carry Initialization, which enables closed-loop motor control by continuously carrying the ongoing recurrent state across successive actions. Building on this formulation, we devised a learning rule termed Feedback-Spike Nudging (FSN), in which teaching signals are delivered as spikes through a feedback pathway. Synaptic changes depend only on presynaptic activity and the feedback-evoked change in postsynaptic firing, consistent with biological mechanisms of synaptic plasticity in motor cortex. The resulting networks learned continuous motor control through both supervised and reinforcement learning, which could be mapped respectively onto interactions of motor cortex with the cerebellum and basal ganglia. Moreover, the FSN learning rule was consistent with experimental observations of structural spine plasticity in mouse motor cortex. Thus, our study lays a foundation for recurrent spiking models of motor cortex that are biologically grounded at both the functional and anatomical levels.

neuroscience↗

bioRxiv2026

Thalamocortical credit routing from the basal ganglia and cerebellum: anatomical constraints and circuit principles

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.

neuroscience↗

bioRxiv2025

Ground tilt representation in the rodent cerebral cortex

Information about ground tilt at ones location is indispensable for integrating the body with its surrounding environment. However, how ground-tilt information, together with movement and posture information, is represented in the cerebral cortex remains unclear. To address this issue, we developed a new six-degree-of-freedom platform that can rapidly and precisely impart ground motion to head-fixed mice. Using the Diesel2p large field-of-view two-photon mesoscope, we performed calcium imaging to simultaneously capture activity dynamics across broad cortical areas from the frontal to the parietal cortex while driving the platform through various motion patterns. We found a core area for the ground-tilt information at the parietal cortex, although tilt-direction neurons were found across the dorsal cortex. Neuronal populations preserved tilt-direction information within a common low-dimensional manifold across three distinct motion conditions, indicating a condition-invariant tilt code. Furthermore, to investigate shared dynamics across multiple cortical areas, we introduced broadcast- subspace analysis as an extension of communication subspace analysis. This extended framework revealed that tilt information occupied the second principal shared axis, following movement information. These results reveal the core cortical area for processing tilt information, its universal format, and its sharing across the dorsal cortex, thereby providing a framework for cortical information processing that integrates movement, posture, and environment.

neuroscience↗

bioRxiv2025

Independence and Coherence in Temporal Sequence Computation across the Fronto-Parietal Network

Time processing requires distributed and coordinated cortical dynamics. Flexible yet robust temporal representations can arise from two distinct computational modes: a coherence mode, where multiple cortical areas hold the same elapsed-time estimate, and an independence mode, where each area maintains its own local estimate. However, how the brain switches between these modes has remained unknown. Using mesoscale two-photon calcium imaging, we simultaneously recorded neuronal populations in the secondary motor cortex (M2) and posterior parietal cortex (PPC) of mice performing a novel alternating-interval timing task. Both areas encoded elapsed time through similar high-dimensional sequential activity. Decoding analyses revealed that the fronto-parietal network has both independent and coherent temporal codes. Communication-subspace analysis showed that temporal information was distributed across multiple low-variance subspaces, whereas the largest subspace preferentially encoded behaviour. A twin recurrent neural network (RNN) model with sparse inter-RNN connections and shared high-variance noise reproduced these experimental findings. Moreover, perturbations applied along the dominant shared subspace paradoxically enhanced independence between the two networks. Through a mathematical formalization based on the local Lyapunov exponents, we uncovered how perturbations along different subspaces selectively evoke either independent or coherent communication mode. Together, these results reveal a principle by which fronto-parietal circuits achieve robust yet flexible computation through the interplay of sparse coupling and shared global fluctuations.

neuroscience↗

bioRxiv2025

Open-source modular FPGA system for two-photon mesoscope enabling multi-layer, multi-depth neural activity recording and lifetime imaging

Large field-of-view (FOV) two-photon microscopy makes it possible to record a large number of neural activities from multiple brain regions simultaneously. However, the larger the field of view, the longer it takes to scan the entire FOV. To increase imaging speed, we have developed open-source software to digitize analogue signals from a photomultiplier tube using a field-programmable gate array (FPGA) at a rate of 3.2 GS/s. By combining this with a newly developed a circular delay-path module for a custom two-photon mesoscope (Diesel2p), we succeeded in simultaneous recording of >10,000 neurons from the entire bilateral dorsal cortex at up to four depths. We also demonstrated large FOV lifetime imaging using the same system. Our modular, open-source FPGA system can be readily integrated into any type of two-photon microscope and will accelerate the biomedical application of multi-scale two-photon imaging in a wide range of pathophysiological investigations.

neuroscience↗

bioRxiv2024

Modulation of hippocampal sharp-wave ripples by behavioral states and body movements in head-fixed rodents

Hippocampal sharp-wave ripples (SWRs) are critical events implicated in memory consolidation, planning, and the reactivation of recent experiences. Under freely moving conditions, a well-established dichotomy exists: hippocampal networks predominantly generate theta oscillations during periods of reward pursuit (preparatory behaviors) and exhibit pronounced SWR activity once the reward is achieved (consummatory behaviors). Here, we examined how SWRs are modulated by reward delivery and subtle movements in head-fixed rats. Contrary to the canonical view established in freely moving settings, we found that the dominant and more enduring effect was a sustained suppression of SWR activity immediately following water delivery. Moreover, even minor, localized movements (such as whisking or body adjustments) decreased SWR occurrence, demonstrating that hippocampal ripple generation is highly sensitive to motor engagement, irrespective of reward timing. Such movement-induced suppression of ripples persisted during both sleep-like states and quiet wakefulness, suggesting that while large-scale brain states modulate the overall likelihood of SWR generation, local motor-related influences exert a state-independent inhibitory effect on hippocampal ripples. Our results show that SWR modulation by behavioral states and body movements is more context-dependent than previously appreciated.

neuroscience↗

bioRxiv2024

Hippocampal sharp-wave ripples decrease during physical actions including consummatory behavior in immobile rodents

Hippocampal sharp-wave ripples (SWRs) are intermittent, fast synchronous oscillations that play a pivotal role in memory formation. It has been well-established that SWRs occur during "consummatory behaviors", e.g., eating or drinking a reward for correct action. However, most of typical behavioral experiments using freely moving rodents have not rigorously distinguished between the act of eating/drinking (regardless of consummation or consumption) from stopping locomotion (immobility). Therefore, in this study, we investigated the occurrence of SWRs during a reward-seeking action and subsequent consummatory reward licking in constantly immobile rats maintained under head fixation and body covering. Immobile rats performed a pedal hold-release action that was rewarded with water every other time (false and true consummation). Unexpectedly, the SWRs remarkably decreased during reward licking as well as pedal release action. Unlearned rats also showed a similar SWR decrease during water licking. Conversely, SWRs gradually increased during the pedal hold period, which was enhanced by reward expectation. A cluster of hippocampal neurons responded to cue/pedal release and reward, as previously shown. Some other clusters exhibited spike activity changes similar to the SWR occurrence, i.e., decreasing during the pedal release action and reward licking, and enhanced by reward expectation during pedal hold period. These task event-responsive neurons and SWR-like neurons displayed stronger spiking synchrony with SWRs than task-unrelated neurons. These findings suggest that the hippocampus generates SWRs, which may associate action with outcome, in "relative immobility" (action pauses) rather than specific consummation or consumption. Significance StatementTo clarify the characteristics of hippocampal sharp-wave ripples (SWRs), we analyzed the SWRs occurring during operant task performance in immobile rats under both head fixation and body covering. First, we found that SWRs decreased when they licked and drank water, conflicting with the theory that SWRs occur in consummatory behavior. Second, hippocampal neurons showed different task-related activities, particularly those that resembled SWR occurrences or conveyed specific signals on task events. Third, these task-related neurons displayed strong synchronous discharges during SWRs in task-engaged periods. These findings may explain the neuronal mechanisms underlying the association between an action and its outcome.

neuroscience↗

bioRxiv2024

Dynamic integration of cortical activity in the deep layer of the anterolateral superior colliculus

The superior colliculus (SC) receives inputs from various brain regions in a layer- and radial location-specific manner, but whether the SC exhibits location-specific dynamics remains unclear. To address this issue, we recorded the spiking activity of single SC neurons while photoactivating cortical areas in awake head-fixed Thy1-ChR2 rats. We classified 309 neurons that responded significantly into 8 clusters according to the response dynamics. Among them, neurons with monophasic excitatory responses (7-12 ms latency) that returned to baseline within 20 ms were commonly observed in the optic and intermediate gray layers of centromedial and centrolateral SC. In contrast, neurons with complex polyphasic responses were commonly observed in the deep layers of the anterolateral SC. Cross-correlation analysis suggested that the complex pattern could be only partly explained by an internal circuit of the deep gray layer. Our results indicate that medial to centrolateral SC neurons simply relay cortical activity, whereas neurons in the deep layers of the anterolateral SC dynamically integrate inputs from the cortex, SNr, CN, and local circuits. These findings suggest a spatial gradient in SC integration, with a division of labor between simple relay circuits and those integrating complex dynamics.

neuroscience↗

bioRxiv2024

Synergistic reinforcement learning by cooperation of the cerebellum and basal ganglia

The cerebral cortex, cerebellum and basal ganglia are essential for flexible learning in mammals. Although traditionally thought to operate under different learning rules, recent evidence suggests that both the basal ganglia and the cerebellum may employ reinforcement learning mechanisms. This raises the question of how these structures coordinate when a common reward prediction error mechanism is active. To address this issue, we first examined output signals from the basal ganglia and cerebellum following the activity of the cerebral cortex. We recorded single-neuron activity from the output regions of the cerebellum and basal ganglia - the cerebellar nuclei (CN) and substantia nigra pars reticulata (SNr) - in both male and female ChR2 transgenic rats. Neurons in the CN and SNr exhibited distinct temporal response patterns; notably, the fast excitatory response in the CN, driven by mossy fiber input, was synchronized with the inhibitory response in the SNr, mediated via the direct pathway. Using these experimental findings together with connectome data, we developed both a semi-realistic spiking network model and a reservoir-based reinforcement learning model. In the latter model, successful learning depended on synaptic plasticity in both the cerebellum and basal ganglia with a temporal precision on the order of 10 ms. Furthermore, cortical {beta}-oscillations enhanced learning and optimal reinforcement learning occurred when the output of cerebellar and basal ganglia signal phase-locked at the frequency of cortical oscillation. Taken together, our results suggest that the coordinated output of the cerebellum and basal ganglia, driven by tightly tuned cortical input, underlies brain-wide synergistic reinforcement learning. Significance StatementThe cerebral cortex, cerebellum, and basal ganglia support learning. Recent research suggests that both the basal ganglia and cerebellum use a similar learning process called reinforcement learning, which involves predicting rewards. To understand how these brain regions work together, we recorded brain activity in rats while photo-stimulating the cerebral cortex. We found that two types of responses in the cerebellum and basal ganglia were synchronized, which might help activate the cerebral cortex. A computer model showed that precise timing of signals from both the cerebellum and basal ganglia is important for learning. This timing was important only when the cerebral cortex worked in a specific frequency range. Our findings suggest that coordinated brain activity enhances learning.

neuroscience↗

bioRxiv2023

Working memory-based and -free reward prediction in a dual dopamine system in the basal ganglia

Animals optimize their actions by predicting and verifying their outcomes (e.g., rewards). Reward prediction often requires working memory (WM)-based information. To elucidate the neural basis of WM-based reward prediction, we compared the activity of dopamine (DA) neurons in the ventral tegmental area (VTA) in an alternate reward condition (WM-based) with that in a random (WM-free) reward condition in rats and mice. Positron emission tomography revealed greater VTA activation in the WM-based than the WM-free condition. Lateral and medial VTA neurons displayed differential electrophysiological spike activities reflecting WM-based and WM-free reward prediction error, respectively. Furthermore, phasic DA release in the dorsal and ventral striatum changed as WM-based and WM-free classical conditioning progressed. Consistent with our WM-based model, reward acquisition caused a DA dip only in the dorsal striatum. Thus a dual DA system processes WM-based and WM-free reward prediction in parallel.

neuroscience↗

bioRxiv2022

Cerebellar climbing fibers convey behavioral information of multiplex modalities and form functional modules

Cerebellar climbing fibers (CFs) convey sensorimotor information and their errors, which are used for motor control and learning. Furthermore, they represent reward-related information. Despite such functional diversity of CF signals, it is still unclear whether each CF conveys the information of single or multiple modalities and how the CFs conveying different information are distributed over the cerebellar cortex. We performed two-photon calcium imaging from cerebellar Purkinje cells (PCs) in mice engaged in a voluntary forelimb lever-pull task and demonstrated that CF responses in 93% of PCs could be explained by the combination of multiple behavioral variables, such as lever movement, licking, and reward delivery. Neighboring PCs exhibited similar CF response properties, formed functional clusters, and shared noise fluctuations of responses. Taken together, individual CFs convey behavioral information on multiplex variables and are spatially organized into the functional modules of the cerebellar cortex.

neuroscience↗

bioRxiv2022

Receptor-like kinases FER and SRK mediate Lotus japonicus freezing tolerance and climate adaptation

Many plant species have succeeded in colonizing a wide range of diverse climates through local adaptation, but the underlying molecular genetics remain obscure. We previously found that winter survival was a direct target of selection during colonization of Japan by the perennial legume Lotus japonicus and identified associated candidate genes. Here, we show that two of these, the FERONIA-receptor like kinase (LjFER) and a novel S-receptor-like kinase (LjSRK) are required for non-acclimated freezing tolerance and show haplotype-dependent cold-responsive expression. Our work demonstrates that recruiting a conserved growth regulator, FER, and a novel receptor-like kinase, SRK, into the set of cold-responsive genes has contributed to freezing tolerance and local climate adaptation in L. japonicus, offering new functional genetic insight into perennial herb evolution.

plant biology↗

bioRxiv2022

Hippocampal CA1 represents action and reward events instantly compared to the superficial and deep layers of the lateral entorhinal cortex

The entorhinal cortex (EC) is the main interface between the hippocampus and the neocortex. The EC plays a critical role in learning and memory. We investigated the neuronal representation of behavioral events during operant learning in the hippocampal-entorhinal circuit of head-fixed rats. Both CA1 and lateral entorhinal cortex (LEC) neurons develop task-related activities after learning. Among diverse task-related activities, we compared the transient peak activities that represent action and reward and found a distinct difference in the timing of behavioral event representation between CA1 and LEC. CA1 represents action and reward events in close to real-time, whereas both the superficial and deep layers of the LEC showed delayed representation of those events. Our results suggest that subpopulations exist within which CA1 and LEC neurons process the information in a different order from the anatomically defined hippocampal-entorhinal circuit.

neuroscience↗

bioRxiv2021

Automated and parallelized spike collision tests to identify spike signal projections

The spike collision test is a highly reliable technique to identify axonal projection of a neuron recorded electrophysiologically for investigating functional spike information among brain areas. It is potentially applicable to more neuronal projections by combining multi-channel recording with optogenetic stimulation. Yet, it remains inefficient and laborious because an experimenter must visually select spikes in every channel and manually repeat spike collision tests for each neuron serially. Here, we established a novel technique to automatically perform spike collision tests for all channels in parallel (Multi-Linc analysis), employing two distinct protocols implemented in a multi-channel real-time processing system. The rat cortical neurons identified with this technique displayed physiological spike features consistent with excitatory projection neurons. Their antidromic spikes were similar in shape but slightly larger in amplitude compared with spontaneous spikes. In addition, we demonstrated simultaneous identification of reciprocal or bifurcating projections among cortical areas. Thus, our Multi-Linc analysis will be a powerful research approach to elucidate interareal spike communication.

neuroscience↗

bioRxiv2018

Extreme genetic signatures of local adaptation during plant colonization

Colonization of new habitats is expected to require genetic adaptations to overcome environmental challenges. Here we use full genome re-sequencing and extensive common garden experiments to investigate demographic and selective processes associated with the recent colonization of Japan by Lotus japonicus. We carefully track the colonization process where L. japonicus gradually spread from subtropical conditions to much colder climates in northern Japan. We characterize the loss of diversity during this process and identify genomic regions with extreme genetic differentiation. Next, we perform population structure-corrected association mapping of phenotypic traits measured in a common garden and discover a number of genome-wide significant associations. Contrasting these analyses, we find that there is a strong concordance between phenotypic variation and extreme differentiation for overwintering and flowering time traits. Our results provide evidence that these traits were direct targets of selection by local adaptation during the colonization process and point to associated candidate genes.

evolutionary biology↗