Poster Session

Like human infants learning to speak, juvenile male zebra finches learn to sing by forming auditory tutor’s songs (TS) memories and then vocally matching to establish their stereotyped own song in the developmental song learning periods. However, neural mechanisms supporting auditory memory-guided vocal motor learning have remained elusive. Recently, we revealed neuronal projections into the song premotor region, HVC, from the neurons responsive to TS in the auditory forebrain area, caudomedial nidopallium (NCM) in the juveniles within song learning period (~60 days post hatch (DPH)). As only few the NCM-HVC projections were observed in the juvenile at the end of it (~90 DPH) in contrast, those suggest transient NCM-HVC projections for developmental song learning and subsequent crystallization (Louder et al., 2024). Here, we performed single-nucleus RNA sequencing (snRNA-seq) analysis of the auditory memory NCM neurons into HVC to elucidate the underlying molecular mechanism of transient projections during developmental song learning period. The NCM neurons responding to TS or HVC neurons responding to a bird’s own song (BOS) were induced to express EYFP by using the AAV-cFos-TetON-EYFP-PEST in the juveniles at different developmental stage. Then the nuclei of NCM or HVC neurons were collected, and the gene expression profiles were compared between EYFP-positive song responding neurons and other EYFP-negative neurons within the NCM or HVC. Post mRNA-sequencing analysis using Seurat revealed several differentially expressed genes (DEGs) in the song responsive EYFP mRNA positive neurons comparing to negative neurons within the NCM or HVC at each age, including; (1) upregulation of vesicle trafficking molecule Rab33B in NCM at 60 DPH when NCM neurons projecting into HVC (2) downregulation of synapse adhesion molecule Teneurin-3 (TENM3) in NCM at 70 and 80 DPH, just before NCM-HVC projection disappears (3) downregulation of synapse adhesion molecules, Latrophilin-3 (ADGRL3) and Kirrel3 in HVC at 75 DPH. Taken together those suggest that the changes in the expression of synapse adhesion and membrane trafficking molecules mediate dynamic remodeling of intercortical auditory-motor projections in the developmental sensorimotor learning period.

Zebra finches, the most commonly used model of songbirds, learn songs from a tutor (normally their father) effectively through social communications, and less efficiently from song playbacks. In zebra finches, only males learn to sing by forming a memory of tutor’s songs (sensory phase) and then vocally matching it (sensory-motor phase). It has been suggested that the caudomedial nidopallium (NCM), a brain area analogous to the mammalian higher auditory cortex, is the neural substrate of tutor’s song memories. Recently we reported inhibition of the inputs from the locus coeruleus (LC), the noradrenergic command center of the brain, in NCM of male juveniles during vocal communication with a tutor impaired song learning, suggesting involvement of noradrenergic signals in song learning (Katic et al., 2022). To further investigate this we tracked noradrenaline release in the NCM during tutor song exposure using a fluorescence noradrenaline sensor (GRABNE) and fiber photometry. At early stage of song learning (58 days post hatch (dph)) a rise of fluorescence signal consecutive to tutor song was recorded 4 / 5 times (80%). However, this response diminished as learning progressed (62 dph: 13/32 times, 41%; 71 dph: 5/24 times, 21%; 85 dph: 4/28 times, 14%) suggesting a regulation of noradrenergic signaling.
We also assessed whether social isolation from a tutor, known to delay the closure of the sensory phase, alters adrenergic receptor expression in the NCM. Using in situ hybridization, we examined expression of α2B, β2 and α1A adrenergic receptors in NCM of juveniles in song learning period (~60 dph). Isolated birds showed elevated expression of β2 and α1A receptors and reduced expression compared to normally reared juveniles. Together, these results suggest that noradrenergic signals in the NCM regulate zebra finch song learning and further shape the time window of song learning critical period.

How does cortex learn to perform complex functions? Classical predictive coding models (Rao & Ballard, 1999) propose that learning is based on mismatches between expected and actual stimuli. Similarly, the error backpropagation algorithm, underpinning the deep learning revolution, enables efficient learning through neuron-specific error signals. This motivates theories of efficient learning in cortex based on the propagation of mismatch/error signals.

However, such bio-plausible models of learning in cortex typically operate on high levels of abstraction.Common issues include: oversimplified neuron models, unexplained information sharing between distant synapses and neurons, biologically implausible network architectures, and phased (interrupted) learning.

We introduce a biologically motivated, multi-area circuit model of cortex. We show that such networks can learn complex tasks using multiple cortical areas, while systematically addressing the aforementioned issues of bio-plausibility. Finally, we discuss ongoing experiments to uncover how cortex learns from errors.

Q-neuron-induced hypothermia and hypometabolism (QIH) is an extreme yet reversible metabolic shift. Surprisingly, previously formed fear memories are preserved during and retrieved successfully post-QIH. There is substantial evidence from past studies that astrocytes stabilize synapses. Our data showed enhanced astrocytic cFos and astrocyte-spine interface (ASI) during QIH indicating potential involvement of astrocytes. Furthermore, astrocyte inhibition by iβark impaired fear memory retention during QIH. Based on these preliminary observations, we hypothesize that astrocytes may play a critical role in memory maintenance during QIH. To further clarify the underlying mechanisms, we aim to use viral based approaches to manipulate astrocyte activity and astrocyte-neuron interaction. In addition, we will employ CDeep3M, an advanced deep learning algorithm and Imaris-based reconstructions of SBF-SEM images to uncover changes in astrocytic morphology and synaptic ultrastructure during QIH. Together, our study aims to provide a comprehensive explanation of how astrocytes aid in memory maintenance during QIH and potentially uncover astrocytic pathways that could be targeted to treat memory disorders.

Electrophysiological properties of cells underlie the fundamental mechanisms of the brain. Although astrocytes have typically been considered not electrically excitable, recent studies show depolarization of astrocytes induced by neural stimulation.
Interestingly, astrocytic depolarization is induced within the periphery of cortical somatosensory astrocytes, proposed to be at contact sites between neurons and astrocytes. As depolarization alters astrocyte neurotransmitter uptake and synaptic efficacy, we used computational models to examine various scenarios that can cause depolarization.
In our model, we explore various hypotheses related to potassium and neurotransmitter transporters and the possibility of depolarization by astrocytic neurotransmitter receptors. We find that astrocyte depolarization can occur under two scenarios, where depolarization is induced by local spillover and high-frequency stimulation, as well as depolarization at the single synapse level as a response triggered by neurotransmitter receptors.
Our findings suggest a new mechanism for how astrocytes can interact with learning and memory in the brain.

Switching walking direction according to environmental demands is an essential function for locomotion in mammal. While the neural circuits mediating forward walking —from the brainstem to the spinal cord— have been extensively studied, the mechanisms underlying backward walking remain unexplored. Understanding these mechanisms represents an important frontier in the study of locomotor control. In this study, we focus on a population of backward-walk-inducing descending neurons (BDNs) that we recently identified. We aim to (1) map their input and output circuits, (2) characterize their activity dynamics during backward walking, and (3) elucidate their physiological roles in locomotor behavior. Through these approaches, we seek to uncover the broader neural principles governing motor control in mammals.

Touch is one of the most fundamental senses connecting the external world to the body. It plays a crucial role not only in perceiving the environment and avoiding danger, but also in forming social bonds and emotional experiences. Among various types of touch, gentle or affective touch—such as caress-like touch, embracing, or social grooming—has been shown to be deeply involved in emotional regulation and sociality, particularly in mammals, where it is tightly linked to healthy development and psychological well-being.
Recent studies have demonstrated that gentle touch is primarily transmitted by low-threshold, unmyelinated sensory fibers known as C-tactile (CT) afferents (Neuron, 2014). These fibers respond optimally to slow, caress-like stimuli delivered at approximately 1–10 cm/s. Once activated, the signals are conveyed from the skin to the spinal cord and then to the brain. In humans, fMRI studies have repeatedly shown that such affective touch activates the posterior insular cortex, a region involved in emotional processing (J. Neurosci., 2011).
Affective touch plays a vital role in early development, as seen in maternal-infant skin-to-skin contact or social grooming between peers (social grooming). These behaviors contribute to attachment formation, stress reduction, learning ability, and the development of social behavior (Neuron, 2022). Conversely, individuals who experience insufficient tactile stimulation during early life may exhibit increased anxiety, and deficits in social interaction.
Animal models, especially rodents, provide valuable insight into the neural basis of gentle touch. In mice, artificial brushing has been shown to activate the brain’s reward circuits and induce conditioned place preference (CPP), suggesting that gentle touch possesses inherent rewarding properties (Science, 2022). In our research, we utilized immediate early gene expression (Fos) to map brain-wide activity following gentle touch stimulation. This approach revealed activation in several emotion- and sensation-related regions, including the primary somatosensory cortex (S1), posterior insular cortex (pIC), and medial amygdala (MeA).
These findings collectively highlight the critical role of affective touch in emotional and social brain functions and lay the foundation for future studies into its relevance in neurodevelopmental and psychiatric conditions.

In primates, neurons giving rise to the corticospinal tract (CST) are distributed in several motor-related areas of the frontal lobe, such as the primary motor cortex (M1), the supplementary motor area (SMA), and the dorsal and ventral divisions of the premotor cortex (PMd, PMv). Recently, we have analyzed the morphology of basal dendrites of putative CST neurons in macaque monkeys and shown that dendrite morphology varies across motor-related areas. Here, we investigated the alterations in basal dendrite morphology of CST neurons after spinal cord injury (SCI). In our SCI model, both the intersection number and the spine density of basal dendrites were highly decreased throughout the motor-related areas.
Notably, these events were less prominent for the PMd than for the M1, SMA, and PMv. When we further compared the density changes post-SCI of the filopodia-, thin-, stubby-, and mushroom-type spines in individual areas, it was found that the density of filopodia-type (immature) spines was increased for all the areas, whereas the other types of spines exhibited their density reductions. Our findings show that SCI caused the morphology change in putative CST neurons associated with impaired manual dexterity not only in the M1 but also in other motor-related areas.

Synaptic vesicle release plays a critical role for neural transmission and synaptic plasticity.  Two different release modes have been observed: synchronous release (SR) with voltage stimuli and asynchronous release (AR) after the stimulus. We investigate the mechanism underlying the different modes of SR and AR, which involves important function of calcium sensors such as the fast sensor of Synaptotagmin 1(Syt1) and the slow sensor of Synaptotagmin 7(Syt7). By characterizing the intrinsic affinity to calcium of Syt1, Syt7 and buffer Calmodulin(CaM), we use the STochastic Engine for Pathway Simulation (STEPS) software to simulate release modes of vesicles from a Readily Releasable Pool (RRP). Our results show that Syt1 has a higher threshold of calcium concentration to trigger vesicle release than Syt7; and in physiologically realistic conditions Syt1 triggers SR and Syt7 triggers AR. Our model demonstrates that slow and fast calcium sensors compete for binding calcium to trigger SR and AR in various calcium dynamics, which can be drastically modulated by buffer CaM. Our model provides a general mechanism for understanding various vesicle releasing kinetics by competition between different types of calcium sensors and buffers.

Spiking microcircuit models exhibit healthy and unhealthy dynamics by simulating single neurons, action potentials and synaptic transmission. This level of detail could be used to identify personalized compensatory mechanisms for brain disorders; however, identifying model parameters that can produce patient data has been challenging. We use neural posterior estimation (NPE), a modern method of simulation-based inference (SBI), to efficiently estimate parameter distributions given dynamics. Using NPE, we showed that the compensatory mechanisms that maintain specific dynamics depend on the underlying epileptogenic cause. This demonstrates that simulators could assist in identifying optimal treatment options.

Phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) plays a crucial role in synaptic vesicle recycling and neurotransmitter release at active zones (AZs) of presynaptic terminals by regulating key endocytic and exocytic proteins. In Down syndrome (DS), a trisomy of human chromosome 21, orthologous to mouse chromosome 16, leads to the overexpression of numerous genes, including several involved in synaptic function, such as synaptojanin-1, a phosphoinositide phosphatase with both 5- and 4-phosphatase activity. This suggests that the levels of PI(4,5)P₂ and its precursor PI(4)P are reduced in the brain due to accelerated turnover.
Here, we investigated the subcellular distribution of PI(4,5)P₂ in the presynaptic membranes of cerebellar parallel fiber (PF) terminals in Ts65Dn mice, a widely used DS model, using SDS-digested freeze-fracture replica labeling (SDS-FRL) combined with electron microscopy. We mapped PI(4,5)P2 localization at the protoplasmic face (P-face) of presynaptic membranes by immunogold labeling using a specific PI(4,5)P2 probe, the recombinant Pleckstrin-homology (PH) domain of phospholipase δ1 detected with a 5-nm gold-conjugated antibody. Quantitative analysis of the gold particle distribution pattern revealed a significant reduction in PI(4,5)P2 density specifically within the AZs of Ts65Dn mice compared to euploid controls, whereas the density outside AZs remained unchanged. In parallel, patch-clamp recordings from Purkinje cells showed a decreased paired-pulse ratio of excitatory postsynaptic currents (EPSCs) evoked by electrical stimulation of PFs, suggesting increased presynaptic release probability in Ts65Dn mice. These findings demonstrate a selective reduction of PI(4,5)P2 at AZs in DS model mice, supporting the hypothesis that altered phosphoinositide organization contributes to presynaptic dysfunction in Down syndrome.

Synapses are integral to information processing and plasticity in the brain. Their functional properties, such as their strength and plasticity profiles, are determined by their underlying structure and organization. A recent study (Rollenhagen et al.) imaged the structure and vesicle organization of Layers 1-6 presynaptic boutons in the human Temporal Lobe Neocortex (hTLN) using transmission electron microscopy. The data reveal variations in vesicle size and position within each bouton, as well as differences in bouton organization across cortical layers. Building on these data, we develop biophysically detailed, spatially explicit stochastic models of hTLN presynaptic boutons using STEPS (Stochastic Engine for Pathway Simulation) to investigate how neurotransmitter release is shaped by the organization of these boutons. We examine how vesicle size diversity contributes to variability in mEPSC (miniature excitatory postsynaptic current) amplitudes and how vesicle size influences vesicle diffusion and fusion probability. Our study also suggests layer-specific differences in VDCC (voltage dependent calcium channel) expression in hTLN boutons, helping reconcile structural and functional experimental observations.

The claustrum is a thin structure densely connected with the brain. Notably, with the default mode network (DMN) it has both a strong anatomical connectivity and, during resting states, a strong functional connectivity.
Here, we observed the behaviour of the claustrum neurons and the activity of the retrosplenial cortex (RSP, an important component of the DMN) during waiting for a long period for reward. RSP is involved in either spatial-oriented functions by translating between self-perspective and allocentric perspective, or non-spatial functions by comparing perceptual input with memory. Because the claustrum targets non-spatial processing cells in RSP, we hypothesized that the behaviour of the two would be correlated during waiting for reward, when the demand for spatial information is minimum.

To this end, first we recorded with fibre photometry the activity in the RSP of 5 mice. We induced locally either the expression of GCaMP6s in most types of neurons with simultaneous retrograde Cre and AAV9.Syn.FLEX.GCaMP6s injections, or GCaMP8m expression in CaMKII positive neurons with an AAV9.CaMKII.jGCaMP8m injection. Second, in another mouse, we injected retrograde Cre in RSP and AAV9.Syn.FLEX.GCaMP6s in the claustrum, to express GCaMP6s in claustrum neurons projecting to RSP. In this case, only the claustrum was imaged, through an nVoke miniscope (Inscopix), via an implanted prism GRIN lens. All mice performed a freely-behaving, delayed-reward task. The waiting times for reward increased from session to session from 0.5 s to 20 s, some reaching up to 120 s.

The results show a strong decrease in the activity of the RSP as soon as the mouse engages in waiting at the food site. During long waiting times (> 10 s), the deactivation reaches a plateau that is maintained until the food is delivered. Then, when the food is delivered, the activity shortly increases during the consumption of the reward, followed by an even stronger decrease than during waiting for reward. About half of the imaged claustrum cells consistently showed short firing bursts immediately before waiting or during waiting. The peak in the average activity of all the imaged claustrum cells coincides with the moment when the activity in the RSP decreases sharply. A local peak in the average activity was also present after the consumption of the reward.

The results suggest the use of the feedforward inhibition from the claustrum to RSP to switch off the spatial-processing cells and to promote the non-spatial processing cells during waiting and after reward consumption. Future optogenetic manipulations of claustrum neurons will clarify this causal relationship.

The brain utilizes Bayesian inference and reinforcement learning (RL) for decision-making. The "Dual cortical computation hypothesis" posits that the sensory cortex performs Bayesian inference while the motor cortex handles RL. We investigated how these computations are implemented across different layers. We recorded multi-layer neural activity from the primary somatosensory cortex (S1) using miniscope calcium imaging while mice performed a lever push-pull perceptual decision-making task. Behaviorally, we found that prior probability, manipulated via block structure, and likelihood via largeness of resistance affected the mice's choices. Encoding and decoding analyses of neural activity were utilized to investigate representation of Bayesian variables in S1 L2/3 and L5.

Human decision-making is often examined by contrasting arbitrary choices, which lack consequence, with deliberate choices, which are reasoned and meaningful. Electroencephalography (EEG) studies have reported readiness potentials preceding arbitrary but not deliberate choices, suggesting distinct underlying neural mechanisms. We re-analyzed EEG data from a donation-choice task using multiple entropy and complexity measures. Our results revealed clear differences in post-stimulus entropy dynamics between decision types. Deliberate choices exhibited higher entropy and complexity than arbitrary choices. These findings reframe neural “noise” as a functional element of cognition rather than a nuisance variable. Entropy does more than quantify uncertainty in brain signals—it captures how unpredictability is selectively recruited when decisions are meaningful versus arbitrary. Entropy-based measures thus offer promising biomarkers of agential involvement, highlighting the constructive role of neural variability in adaptive behavior and decision-making.

The striatum is a key brain region for the ability to adapt to a changing environment and learning of beneficial relationships between stimuli and positive outcomes, such as food. However, erroneous association acquisition can lead to unhelpful behaviours, and impairments of this cognitive function are hallmarks of several neurological diseases. Therefore understanding how the striatum is involved in associative learning has important implications for addressing these issues.
The dynamic relationship between two neuromodulators, acetylcholine (Ach) and dopamine (DA), in the striatum is implicated in this learning process. However, the ability to illuminate the exact nature of their interactions and how this contributes to learning has been technologically restricted. Recent advancements in genetic biosensors have enabled recording of various neurotransmitters, including Ach and DA, whilst rodents engage in cognitive tasks with appropriate spatiotemporal resolution. Employment of these biosensors is now challenging the previously held beliefs about the nature of Ach/DA interactions and the role in cognition. Cholinergic interneurons (CINs), which release Ach, were shown in vitro to regulate DA release but recent in vivo evidence of this regulation is inconsistent. Most studies have employed fibre photometry to record DA with biosensors to investigate the relationship with Ach, which provides a regional average but masks any variability in response. By instead combining dLight (DA biosensor) with 2-photon imaging it is possible to explore regional heterogeneity in dorsomedial cue-evoked DA release during Pavlovian and instrumental conditioning in mice, and how chemogenetic inhibition of CINs impacts this release and associated behavioural readouts. Preliminary data from this project indicates heterogenous changes in cue-evoked DA positively correlated with time in a conditioned mouse.
This ongoing work will provide both detailed spatiotemporal representations of DA release during two types of associative learning and contribute to clarification around potential DA release modulation by CINs during behaviour.

The cephalopod peduncle lobe has been analogized to the vertebrate cerebellum, based on similar anatomy and behavioral effects of lesions. A common view on cerebellar computation is that it is implementing internal models for motor control. Do these facts imply that the peduncle lobe implements an internal model? To test this idea, we are conducting high-speed, markerless 3D tracking of squid (Sepiotheutis lessoniana) during various behavior. We seek to record the signatures of internal model-based predictions through quantitative behavioral analysis. We are combining these experiments with peduncle lobe lesions. Through profiling internal models and their basis in cephalopods we seek to discover general principles governing the implementation of sensory-motor predictions.

Understanding how meaning arises from brain activity remains one of the biggest frontiers in neuroscience. Traditional correlation-based models identify co-activation but cannot describe the causal geometry underlying how distributed neural processes evolve through time. Our goal is to develop an interpretable and dynamical framework for decoding semantic primaries—core conceptual dimensions such as emotion, vision, and time—directly from naturalistic brain imaging. We apply Manifold Dimension Expansion (MDE), a recent extension of Empirical Dynamic Modeling (EDM), to brain recordings from the Narratives dataset (ds003020), where participants listened to real-world stories. After Schaefer-400 parcellation, regional activity was transformed into a comprehensive ROI–lag matrix, capturing time-lagged interactions across the cortex. This full matrix is passed directly into MDE. Within MDE, each ROI–lag is evaluated using nonlinear state-space forecasting (simplex projection) to assess its ability to predict semantic category trajectories derived from English1000 and word2vec embeddings. MDE then performs a deterministic, validation-driven greedy expansion, iteratively adding the ROI–lag component that most improves out-of-sample semantic forecasting. The final model is evaluated only on a held-out test segment to prevent leakage. The resulting manifold reveals how spatially distributed and temporally offset neural signals cooperate nonlinearly to encode semantic primaries. Unlike contemporary deep learning decoders—powerful but opaque—EDM–MDE provides causal transparency: every predictive element corresponds to a measurable brain region and lag, rather than a hidden parameter. These results demonstrate that meaning can be forecast directly from brain dynamics using interpretable, data-driven nonlinear modeling, offering insight not only into what the brain represents, but how those representations evolve.

To understand consciousness, it is necessary to integrate diverse methods, theories and perspectives from neuroscience and cognitive science. In order to organise the rapidly expanding field of neuroscience of consciousness, we created a map that charts conceptual domains, methodological approaches and theoretical positions, thereby clarifying how different research programmes intersect. Incorporating multiple perspectives, from first-person analyses to systems neuroscience and computational modelling, provides a fuller picture. Emerging trends, including the creation of systematic catalogues of artificial and biological models, highlight a shift towards comparative and integrative approaches. We propose a topological model that disentangles the dimensions of consciousness, intelligence, and agency in order to demonstrate the need to reinstate conscious efficacy in the neuroscience of consciousness.

Our previous research revealed a causal relationship between dorsal raphe serotonin neural activity and operant waiting for future rewards by showing optogenetic activation and inhibition of serotonin neurons prolonged and shortened waiting time for future rewards, respectively (Miyazaki et al., 2014, Taira et al., 2024). We also found that serotonin effect in promoting waiting is maximized by both high reward probability and high reward timing uncertainty (Miyazaki et al., 2018). We further proposed a Bayesian decision model of waiting which assumes serotonin activation increases the prior reward probability to reproduce the major features of the experimental results (Miyazaki et al., 2018, 2020). However, there was no direct evidence serotonin neural activity was modulated by reward probability.
Tph2- GCaMP6 transgenic Mice (n = 7) in which GCaMP6 is selectively expressed in serotonin neurons were trained to perform a sequential tone-food waiting task (Miyazaki et al., 2014, 2018) that required them in one trial to wait for a delayed tone (tone delay: 0.3 s, tone duration: 0.5 s) at tone site and then to wait for a delayed food (reward delay: 3 s) at reward site. The sequential tone-food waiting task lasted until the mouse had completed 40 trials. An optical fiber (400 m diameter) was implanted into the DRN and serotonergic neural activity was recorded by fiber photometry. We prepared four reward probability tests (100, 75, 50, and 25%) and associated with four tones (8 kHz, 2.1 kHz, white noise, and 4.5 kHz), respectively. In one test, the reward probability was constant. We focused on serotonergic neural activity during operant waiting for delayed reward. dF/F was the highest in the 100% test 23.9 ± 2.1% and gradually decreased 19.5 ± 1.9%, 16.7 ± 1.9%, and 14.2 ± 1.8% in the 75%, 50%, and 25% tests, respectively. When the reward amount presented in reward trial was changed in the 25% test, waiting time in the 25% three-pellets test (5.1 ± 0.1 s) was significantly longer than that in the 25% one-pellet test (4.8 ± 0.1 s). This result indicates that waiting behavior for future rewards is affected by reward amount. We found that there is no significant difference in serotonin neural activity during waiting between the 25% three-pellets test (14.1 ± 1.7 %) and the 25% one-pellet test (14.2 ± 1.8%). These results show dorsal raphe serotonin neurons encode probability not value of future rewards and serotonin response would be used to support flexible behavior.

We have developed a marker-based 3D motion capture system for high-precision quantification of locomotion in mice. This method allows measurement of fine 3D movement trajectories such as step kinematics during various behaviors with submillimeter accuracy. In our previous studies, we observed that a low dose of a synthetic agonist of cannabinoid CB1 and CB2 receptors, CP55,940 (0.3 mg/kg), causes a small reduction in general locomotion during open-field exploration but does not inhibit locomotion during vertical climbing. However, we observed that it induces significant changes in step kinematics, which were more pronounced during climbing. In the present study, we aimed to test whether the step kinematics of mice is affected during climbing and high-speed treadmill running under treatment with CP55,940 and tremor-inducing harmaline.
We used adult male C57BL/6 mice in a within-subject, randomized design (n=6-10). We recorded voluntary climbing behavior on a spoked mesh wheel, as well as high-speed treadmill running in mice treated with vehicle (1:1:18 EtOH, Kolliphor, saline), a low dose of CP55,940 (0.3 mg/kg), or harmaline (20 mg/kg, i.p.) 30 min before recordings. Mice were implanted with permanent markers located on the hips, lumbar spine, shoulder blades, hindlimb knees, and ankles. A high-speed, high-resolution 3D motion capture system (Qualisys, Sweden) was used to track 3D trajectories and the velocity of markers during voluntary locomotory tasks: vertical climbing and running on a treadmill. On treadmill mice were running at speeds ranging from 1 m/min to 40 m/min, doubled with each trial until failure.
During climbing CP55,940 did not reduce general locomotion but affected step kinematics decreasing step height and speed which is consistent with our previous observations. Harmaline, at 20 mg/kg induces strong whole-body tremor and was found to significantly inhibit locomotion during climbing and reduce the total number of steps. Harmaline also significantly inhibited the ability of mice to run on a treadmill. Interestingly CP55,940 also reduced the ability of mice to run on the treadmill despite having no effect on climbing. The maximum speed achieved on the treadmill was significantly decreased by harmaline and CP55,940 by >10 m/min on average. During treadmill running, ankle swings had significantly higher mean and maximum speeds than during vertical climbing or exploration tasks. Moreover, CP55,940 reduced the height of the ankle swing during treadmill running, similarly as it does during climbing.
In conclusion, our results demonstrate the utility of marker-based 3D motion capture in accurate and precise measurement of high-speed fine kinematics such as running or tremor in mice. This innovation will help quantify motor impairments and disruption of kinematics in mouse models.

Astrocytes play a crucial role in the brain. They exihibit dynamic fluctuations in intracellular Ca2+ concentration with a duration that varies from minutes to hundreds of milliseconds. The fastest events have recently attracted increasing attention as they are on the same time scale as neuronal activity. Here, we introduce several of our methodologies for investigating astrocytic Ca2+ dynamics in behaving mice as well as complementary examinations using fixed brain tissue.

RNA based-gene switches are emerging techniques for on-demand protein production in next-generation gene therapy and mammalian synthetic biology. However, few user-friendly tools for highly controlled gene expression are approved for clinical use. Here, we describe a rapid optimization platform for gene switch through a large-scale switch mutant library and massively parallel switch enrichment experiment. Our platform enables the rapid evaluation of over 10,000 gene switches variants in human cells, resulting in efficiently identifying the best-performing switch design within 3 months. We applied our platform to the development of an RNA-based gene switch (riboswitch) that recognizes clinically safe small molecule drugs and provides dose-dependent, temporally controlled therapeutic gene expression. The identified gene switches showed extremely sensitive drug response with an EC50 of ~0.24 μM, outperforming the recently reporteddrug-responsive RNA-based switches (EC50 of ~1.1 μM; Luo et al., Nat Biotechnol, 2024). This technology may enable precise, dose-dependent gene regulation through clinically safe drugs, offering unprecedented potential for next-generation gene therapy with enhanced safety.

STEPS is a voxel-based reaction-diffusion simulator that we develop to simulate subcellular to whole-cell scale molecular neuron models, within realist geometries captured by tetrahedral meshes. In the past we added membrane currents and voltage calculations within a parallel implementation increasing the scope of models that could be simulated to molecular detail.
More recently, we added parallel support for synaptic vesicles, exocytosis, endocytosis, active transport and other new features, alongside the regular reaction-diffusion framework. This enables simulation of the complete synaptic vesicle cycle to molecular detail, and I will describe work published this year in which we apply the software to a hippocampal synapse and investigate recycling and reserve vesicle pool usage at different firing frequencies.
I will also describe unpublished work in which we investigate calcium activation of BK-type potassium channels of different unitary conductances, within the physiological ranges reported across different animal species. We find that activation of the low unitary conductance channels found in invertebrates is dampened by a calcium buffering effect whereas higher unitary conductance channels found in mammals and birds activate more strongly, allowing the channel to repolarise sodium action potentials and leading to more reliable spike timing in these species.

STEPS (steps.sourceforge.net) is a modeling and simulation software package that allows the simulation of stochastic neuronal reaction-diffusion processes on 3D realistic tetrahedral meshes [6].
At its core, STEPS uses Gillespie's stochastic simulation algorithm (SSA) [3] to sample stochastic trajectories from the reaction-diffusion master equation. Although very accurate, this algorithm sequentially samples firings from reaction channels and can thus become particularly slow when chemical reactions are very frequent or when large populations of chemical species are involved.
To mitigate this issue, various approximate methods have been devised such as Tau-leaping [4] and R-leaping [1]. These methods allow for the simultaneous firing of multiple reaction channels while still maintaining an acceptable level of accuracy. Most implementations of these methods are however restricted to single core simulations, which considerably restricts the spatial scales that can be simulated.
Here we present a multi-core implementation in STEPS of an approximate method that allows for substantial speedup of simulations. Our method is implemented within the operator-splitting framework, in which reactions and diffusion events are computed separately. The approximate reaction operator combines a composition-rejection method to R-leaping, while the approximate diffusion operator uses a modified version of tau-leaping that samples net fluxes between tetrahedrons.
We compare the runtime and accuracy of our solution with available reaction-diffusion simulation software like MesoRD [5], URDME [2], and NeuroRD [7].

References:

[1] A. Auger, P. Chatelain, and P. Koumoutsakos. R-leaping: Accelerating the stochastic simulation algorithm by reaction leaps. The Journal of chemical physics, 125(8), 2006.

[2] B. Drawert, S. Engblom, and A. Hellander. URDME: a modular framework for stochastic simulation of reaction-transport processes in complex geometries. BMC systems biology, 6(1):76, 2012.

[3] D. T. Gillespie. Exact stochastic simulation of coupled chemical reactions. The journal of physical chemistry, 81(25):2340-2361, 1977.

[4] D. T. Gillespie. Approximate accelerated stochastic simulation of chemically reacting systems. The Journal of chemical physics, 115(4):1716-1733, 2001.

[5] J. Hattne, D. Fange, and J. Elf. Stochastic reaction-diffusion simulation with MesoRD. Bioinformatics, 21(12):2923-2924, 2005.

[6] I. Hepburn, W. Chen, S. Wils, and E. De Schutter. STEPS: efficient simulation of stochastic reaction-diffusion models in realistic morphologies. BMC systems biology, 6(1):36, 2012.

[7] Z. Jedrzejewski-Szmek and K. T. Blackwell. Asynchronous τ-leaping. The Journal of chemical physics, 144(12), 2016.

Mind-wandering reflects a complex interplay between focused attention and off-task mental states. Various studies have investigated the psychological and systematic mechanisms underlying these shifts. However, the current models have not yet provided an account for the underlying neural mechanisms for autonomous shifts between the two states. Recent works investigated mind-wandering mechanisms using the Predictive Variational Recurrent Neural Network (PV-RNN), a hierarchically organized model rooted in the Free Energy Principle (FEP). The PV-RNN’s dynamic behavior is governed by a meta-level parameter, the meta-prior w, which balances the complexity term against the accuracy term in free energy minimization. While these current computational studies provide critical insights into macroscopic neural mechanisms, the transition from FS to MW has been caused by manual resetting of the meta-prior from a low to a high setting, leaving the mechanism for autonomous FS-MW shifts unexplored.
Motivated by the above, we propose an online adaptive mechanism for w, modulated by the average reconstruction error over a fixed length time window in the past. A simulation experiment is presented to showcase the proposed framework. In particular, using PV-RNN, we trained the model to predict sensory patterns generated by probabilistic transitions among cyclic trajectories. Simulation results demonstrate that autonomous shifts between FS and MW emerged as w switched dynamically: high w enhanced top-down predictions, promoting MW, while low w emphasized bottom-up sensory perception, favoring FS. Finally, this work explores how our experiment results align with existing studies and highlights their potential for future research.

Human infants acquire language and action gradually through development, achieving remarkable generalization capabilities from only a minimal number of learning examples. In contrast, recent large language models require exposure to billions of training tokens to achieve such generalization. What mechanisms underlie such efficient developmental learning in humans? This study addresses this question through simulation experiments in which robots learn to perform various actions corresponding to imperative sentences (e.g., \textit{push red cube}) via trials of self-guided exploration. Our approach integrates the active inference framework with reinforcement learning, enabling curiosity-driven developmental learning. The simulations yielded several important findings:
i) Generalization is drastically improved as the number of compositional elements increases.
ii) Curiosity-driven exploration combined with motor noise substantially outperforms learning without curiosity.
iii) Rote pairing of sentences and actions occurs before the emergence of compositional generalization.
iv) Simpler, prerequisite-like actions emerge earlier in development, while more complex actions involving these prerequisites develop later.
These results shed light into possible mechanisms underlying efficient developmental learning in infants and provide computational parallels to findings in developmental psychology.

Neural activity is fundamentally a dynamical process, yet much of neuroscience still relies on static measures such as average firing rates to decode it. These rate-based approaches neglect the temporal structure inherent in spike trains and the nonlinear dynamics underpinning neuronal computation. Inspired by Takens’ theorem—a foundational result in nonlinear dynamical systems theory—we apply time-delayed embedding techniques to spike timing, enabling the reconstruction of latent dynamical trajectories from single-neuron and population spiking activities. While previous work has shown that individual neuron inter spike interval (ISI) embeddings can reveal intrinsic dynamical structure of single neuron activity, here we extend this framework to multi-neuron ensembles to investigate how collective spike timing could work as a possible population code relates to behavior. As a proof of concept, we analyzed experimental data from trained mice performing maze traverse tasks. We demonstrate that different neurons within the same circuit exhibit diverse, attractor dynamics. Importantly, these reconstructed trajectories not only capture internal dynamical motifs but also predict behaviors with high fidelity. Moreover, we identify transitions in the geometry of these embeddings that correlate with behavioral state changes, suggesting a link between dynamical phase shifts in neural activity and shifts in cognitive or motor states. Our results introduce a principled alternative to spike rate–based analyses, offering a dynamical systems approach to uncover how timing-based neural codes in both single cells and populations map onto behavior. This framework opens new directions for understanding neural computation by bridging single-neuron dynamics, ensemble coordination, and behavior through a common geometrical lens.

Memory consolidation during sleep requires a dynamic balance between the strengthening of memory traces via neural reactivation and synaptic renormalization to prevent capacity overload and facilitate forgetting. We investigate the underlying synaptic mechanisms governing this balance in the hippocampal CA3 region, which is critical for autoassociative memory. The CA3 utilizes a noncanonical, temporally symmetric Spike Timing-Dependent Plasticity (STDP) rule, which robustly drives potentiation regardless of spike order, alongside pronounced activity-dependent synaptic depression (modeled by Tsodyks kinetics) relevant for homeostatic downscaling. This project unifies these neurophysiological observations—symmetric STDP, synaptic depression, and sleep-like memory replay—within a computational autoassociative network model. The core objective is to determine how the interplay between symmetric potentiation and depression influences the network’s capacity to retain encoded information. We will quantify the resulting memory storage capacity and recall robustness under high informational loads, clarifying the combined mechanisms that optimize the stability and persistence of memories during the consolidation phase.

Cephalopod brains undergo continuous growth during the first months of life, yet the mechanisms that sustain this growth while maintaining functional signaling and behavior remain unclear. To investigate this, we exposed hatchling and juvenile Sepioteuthis lessoniana to a proliferative cell marker and sampled whole heads/brains at different time points post-exposure. Brains were cleared and imaged with light sheet microscopy to localize new cells. Although the presence of proliferative populations within the central nervous system (CNS) was uncertain, we found that new cells appear in the CNS over time, migrating through three distinct pathways in S. lessoniana. Within the subesophageal mass, new cells formed layered structures, while in the supraesophageal mass they were more sparsely distributed. A subset of new cells colocalized with neuronal markers, suggesting that they adopt neuronal identities. Ongoing work employs the whole-cell patch clamp technique in in vitro slice preparation to further characterize these cells electrophysiologically and morphologically, and to identify the specific lobes of the central brain where they integrate. Our findings will contribute to understanding how newly born neurons are incorporated into pre-existing circuits and what role postembryonic neurogenesis plays in cephalopod brain development.

In human, mutations on cone phototransduction genes, gnat2, pde6c, cnga3 and cgnb3, cause photopic vision defects called achromatopsia. Among them, gnat2 mutations cause only achromatopsia, whereas pde6c and cnga3/b3 mutations cause cone dystrophy. It is interesting why these mutations show different symptoms, although these genes function in the same cone phototransduction pathway. In zebrafish, similarly, gnat2 mutations cause achromatopsia without cone degeneration, while pde6c mutations cause achromatopsia with cone degeneration.
First, we examined what factor mediates cone degeneration in zebrafish pde6c mutants. A candidate factor is the increase of intracellular concentration of cGMP, Ca2+, or both. We examined intracellular Ca2+ levels in zebrafish photoreceptors by confocal imaging using the GCaMP7a transgenic system. Cytosolic Ca2+ level was significantly elevated in pde6c mutant photoreceptors. On the other hand, immunohistochemical analysis showed that cGMP level was not altered.
Second, we examined whether cnga3 mutation can rescue cone degeneration in pde6c mutants. The elevation of cytosolic Ca2+ level and defects in structural integrity and maintenance of cone photoreceptors in pde6c mutants were significantly rescued by knock-down of cnga3. These data indicate that cone photoreceptor degeneration in the absence of PDE6C activity depends on abnormal elevation of cytosolic Ca2+ concentration, which may be caused by chronic opening of CNGA3. Third, we examined whether gnat2 mutation can rescue cone degeneration in pde6c mutants. We found that gnat2 mutation failed to rescue cone degeneration in pde6c mutants, suggesting that a simple blockade of phototransduction pathway does not induce photoreceptor degeneration, but that absence of PDE6C activity directly causes cone degeneration. These findings advance our understanding on pathological process of human photoreceptor degeneration linked to PDE6C mutations.

Affective touch (A-touch), defined as gentle, emotionally positive tactile stimulation such as stroking or grooming, plays a crucial role in shaping social and emotional development. During early postnatal life, tactile interactions between the mother and offspring provide essential sensory experiences that promote attachment and communication. Recent studies have identified MrgprB4-, Prokr2-, and GPR83-expressing C-LTMR neurons as key mediators of A-touch signaling, projecting from the skin through the spinal cord to higher brain regions involved in emotion and social cognition. However, how A-touch input during development organizes these brain circuits remains poorly understood.
This study aims to elucidate how deprivation of A-touch during critical developmental windows alters the formation of social circuits and subsequent social behavior. Using transgenic mice that allow temporal control of A-touch neuron ablation, tactile inputs will be selectively blocked at defined developmental stages, and behavioral and neural consequences will be assessed longitudinally from infancy to adulthood. Behavioral paradigms include ultrasonic vocalization (USV) and contact-dependent social preference tests, enabling a comprehensive evaluation of communication and affiliative behavior. Complementary brain mapping and neural activity imaging will further identify circuit-level alterations induced by A-touch deprivation.
By combining developmental manipulation with behavioral and system-level circuit analyses, this project will clarify how early tactile experiences sculpt emotional and social brain networks. The findings are expected to provide mechanistic insights into how disruptions of affective touch during development contribute to neurodevelopmental disorders such as autism spectrum disorder and attachment dysfunction.

Migraine is a severe neurological disorder. One of the key mechanisms underlying its development is considered to be central sensitization in the caudal nucleus of the trigeminal nerve. It is known that neuroinflammation can activate the trigeminovascular system. However, the mechanisms by which it contributes to migraine pathogenesis remain unclear. In the present study, we investigated the development of mechanical hypersensitivity and photophobia in migraine associated with induced inflammation, as well as the activity of 5-HT and P2X receptors in peripheral afferents.
Electrophysiological recordings of trigeminal nerve action potentials were performed in hemiskull preparations of rats.
Our results demonstrated that allodynia and photophobia were enhanced under migraine conditions following prenatal administration of lipopolysaccharide (LPS). Furthermore, acute LPS administration increased the electrical activity of trigeminal afferents induced by 5-HT receptor agonists. Modeling migraine under prenatal inflammatory conditions also intensified mast cell degranulation in the meningeal membranes of rats.

Movement is a fundamental feature of our behaviors and significant portion of nervous system contributes to its generation. Descending pathways are known as the circuits which descend motor signals from the brain to the spinal cord to produce motor outputs. Major descending pathways in mammals include the corticospinal tract (CST), the rubrospinal tract (RST), and the reticulospinal tract (ReST). While the CST has been extensively studied, other pathways that have the potential to contribute to recovery after spinal cord injury or stroke remain less understood. In particular, ReST neurons, located in the caudal medulla and projecting axons to the cervical spinal cord, comprise a diverse population of neurons including not only excitatory but also inhibitory projection neurons. However, the specific role they play in regulating movement is not yet fully understood. To clarify its characteristic, we are investigating the functional roles of excitatory and inhibitory neurons of the ReST during skilled reaching movement. Two transgenic mice lines, vglut2-Cre and vgat-Cre, were used to selectively target excitatory and inhibitory neuronal populations of the ReST, respectively. For optogenetic manipulation, AAVrg-hSyn-SIO-stGtACR2 and AAVrg-hSyn-DIO-rsChRmine were injected into the mouse cervical cord C6-C8 to retrograde express excitatory and inhibitory opsins in ReST neurons. Subsequently, optical fiber was implanted into the reticular formation of the medulla to enable activation or inhibition of excitatory (vglut2-positive) and inhibitory (vgat-positive) ReST neurons. The mice utilized in the inhibition experiments were trained to execute reaching movements in a head-fixed position. Behavior analysis is ongoing using high speed cameras. Recorded videos are going to be analyzed by DeepLabCut software.
Furthermore, histology analysis is underway to verify retrograde labeling by AAV injection. We expect to observe behavioral patterns that clarify distinct roles of excitatory and inhibitory ReST neurons in skilled movement generation. This rotation project aims to advance our understanding of how descending motor circuits coordinate reaching behaviors and shed light on pathways potentially relevant for motor recovery after injury.