TOP ＞ Late-Breaking Abstracts

Late-Breaking Abstracts 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-001 Gad2ノックアウトラット脳におけるGAD67の細胞内局在 Intracellular localization of GAD67 in the Gad2 knockout rat brains *柿崎 利和(1)、柳川 右千夫(1) 1. 群馬大学大学院医学系研究科遺伝発達行動学 *Toshikazu Kakizaki(1), Yuchio Yanagawa(1) 1. Department of Developmental Genetics and Behavioral Neuroscience, Gunma University Graduate school of Medicine Keyword: GABA, GAD, Intracellular localization GABA is a major inhibitory neurotransmitter in the adult mammalian CNS. Dysfunction of GABAergic neurotransmission is implicated in several psychiatric disorders including schizophrenia, major depressive disorder and attention-deficit/hyperactivity disorder. GABA is synthesized by two types of glutamate decarboxylase (GAD), GAD65 and GAD67, which are encoded by the independent genes, Gad2 and Gad1, respectively. The brain GABA was undetectable in Gad1/Gad2 double knockout rats or mice. Therefore, GADs are indispensable molecules for GABAergic neurotransmission and can be an important therapeutic target for several psychiatric disorders. It is thought that GAD65 is mainly localized to the synaptic terminals and plays a role in use-dependent synthesis of GABA, while GAD67 is widely distributed throughout the cells and is responsible for the basic requirement for GABA. Furthermore, the GAD65/GAD67 protein ratio (GAD65/67 ratio) in brain is also known to be different among species. Our western blot analyses with anti-GAD65/67 antibody revealed that the mouse and rat GAD65/67 ratio was about 1 and 3, respectively in the cerebral cortex at postnatal week 8. Thus, while almost equal amounts of GADs are contained in mouse cerebral cortices, GAD65 is a major GAD isoform in rat ones. Indeed, the brain GABA content in Gad2 knockout mice was reduced to about half of the wild-type mice. On the other hand, two-thirds of the GABA contents were maintained in the Gad2 knockout rat brains, although there was no difference in the amount of GAD67 proteins between Gad2 knockout and wild-type rat brains. It was reported that the enzymatic activity of GAD67 is increased by the dephosphorylation via calcineurin which is concentrated in the axon terminals. Considering these findings, we hypothesize that GAD67 is accumulated in the axon terminals of GABAergic neurons in the Gad2 knockout rat brains. Here, in order to verify this possibility, we performed immunohistochemical staining with two antibodies against GAD67 and vesicular GABA transporter (VGAT) which is a presynaptic marker of inhibitory neurons. The staining showed a tendency to increase of GAD67+VGAT+ double-positive signals in the Gad2 knockout rat cerebral cortex and hippocampus, compared to those in the wild-type ones. These results imply that GAD67 proteins are accumulated in the presynaptic terminals of GABAergic neurons in the Gad2 knockout rat brains. This alteration might account for the relatively preserved GABA content in the Gad2 knockout rat brains. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-002 cAMP緑色蛍光プローブgCarvi発現神経細胞の基底及び変動cAMP解析 Pharmacological study of basal level and dynamics of cAMP in neurons expressing green fluorescent cAMP probe gCarvi *川田 聖香(1)、齋藤 直人(1) 1. 同志社大学大学院生命医科学研究科 *Seiko Kawata(1), Naoto Saitoh(1) 1. Grad Sch Life Med Sci, Doshisha Univ, Kyoto, Japan Keyword: cAMP, live-cell imaging The second messenger cyclic adenosine monophosphate (cAMP) plays various roles in neuronal functions. Molecular cascades underlying cAMP-mediated signaling pathways are well clarified, but its spatio-temporal profiles of intracellular cAMP remain hidden. We have developed a new genetically encoded monomeric green fluorescent cAMP indicator named “gCarvi”. This probe monitors intracellular cAMP within 0.2–20 µM with a Kd of 2 µM covering of neuronal physiological cAMP range. gCarvi can rapidly detect cAMP with a Kon of 1.38 ± 0.25 × 106 mol−1 s−1 and Koff of 3.31 ± 0.77 s−1. Compared to previously reported Green Up cADDis cAMP in BacMam, gCarvi indicated 2-fold faster rise-time of cAMP elevation stimulated by isoproterenol (a β-adrenergic receptor agonist) in COS-7 cells. gCarvi can be converted to a retiometric probe for intracellular cAMP concentration. mCherry fused gCarvi (ratiometric gCarvi) expressed hippocampal neurons were permeabilized using 30 µM escin and exposed to cAMP at various concentrations (1–100 µM). The basal [cAMP]i estimated from the titration curve was 1.38 ± 0.59 µM. This concentration is slightly below the level required for PKA activation. The basal [cAMP]i in hippocampal neurons decreased to a minimal level by KH7 (a soluble adenylate cyclase inhibitor) application. The linkage between cAMP and Ca2+ signals is important for various neuronal function. For simultaneous monitoring both messengers, gCarvi and jRCaMP1b (a red fluorescent Ca2+ probe) were co-expressed in hippocampal culture neuron. We bath-applied 10 µM forskolin (an adenylate cyclase activator) and 100 µM IBMX (a phosphodiesterase inhibitor), which caused a gradual increase in the somatic gCarvi fluorescence and reached a plateau within 3 min. Ca2+ response induced by the cAMP elevation categorized into 3 groups; increasing both spontaneous spikes and basal Ca2+ level (group1), only increasing basal Ca2+ level (group2), and no clear change (group3). Forskolin/IBMX induced cAMP elevations of group1 and group2 are similar, and larger than that of group3. These results indicate that elevation of intracellular cAMP above a certain level can trigger tonic basal Ca2+ elevation and then additive Ca2+ spikes in hippocampal neurons. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-003 紫色励起光型cAMPプローブの特性とライブイメージング Properties and live-imaging of violet light excitable cAMP probe *山本 優斗(1)、川田 聖香(1)、齋藤 直人(1) 1. 同志社大学大学院生命医科学研究科 *Yuto Yamamoto(1), Seiko Kawata(1), Naoto Saito(1) 1. Grad Sch Life Med Sci , Doshisha Univ , Kyoto , Japan Keyword: cAMP, live-imaging 3’,5’-cyclic adenosine monophosphate (cAMP) is a well-known molecule as a second messenger in intracellular signal transduction e.g. neuronal differentiation, neurite elongation, and synaptic plasticity. The molecules in cAMP cascade are extensively studied, although spatio-temporal features of cAMP in neuron are not well understood. Förster resonance energy transfer (FRET)-based and circularly permutated-fluorescent protein (cpFP)-based cAMP probes are available so far. FRET-based cAMP probes are composed of the derivative of CFP/YFP pair. cpFP-based cAMP probes utilize the variants of cpGFP, cpYFP or cpRFP for intesiometirc detection. To extend the usage of cAMP probe, we newly designed violet light excitable cAMP probe which can easily combine with the other signaling probes or optogenetic tools. We report here the characteristic of the violet excitable cAMP probe. The spectrum analysis indicates violet excitation and green emission. The dose–response curve of the cAMP probe indicates Kd of 9.49 μM and Hill coefficient of 0.89, ensuring one-to-one stoichiometry. The cAMP probe could be expressed in cultured hippocampal neurons using rAAV-SynTetOff vector. Bath-application of forskolin, general transmembrane adenylyl cyclase activator, induced decrease of the probe’s fluorescence, which was consistent with the data of purified protein. Finally, we co-expressed the cAMP probe with biPAC, blue light activatable adenylyl cyclase, as a proof of the concept. From these results, this newly designed cAMP probe can be useful for multi-signaling imaging. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-004 小脳スライスにおけるプルキンエ細胞の単純スパイクの膜電位イメージング Voltage imaging of simple spikes from Purkinje cells in acute cerebellar slice *橋田 英吉(1)、真仁田 聡(1)、喜多村 和郎(1) 1. 山梨大学大学院総合研究部医学域生理学講座　神経生理学教室 *Eikichi Hashida(1), Satoshi Manita(1), Kazuo Kitamura(1) 1. Department of Neurophysiology, Faculty of Medicine, University of Yamanashi Keyword: Genetically-encoded voltage indicator (GEVI), Adeno-associated virus (AAV) Purkinje cells are the only output cells in the cerebellar cortex and are known to produce two types of spikes: simple and complex spikes. The response characteristics of these spikes have been investigated by electrophysiological methods. However, these recording methods are limited in the number of cells recorded, and the fine spatial information is lacking. Calcium imaging allows us to observe the activity of multiple neurons simultaneously with more spatial information. However, individual simple spikes of Purkinje cells cannot be detected using calcium imaging. This is because intracellular calcium elevation by each simple spike is very small in Purkinje cells, and the temporal resolution of calcium imaging is much lower than the simple spike rate up to 100 Hz. The exact timing of the simple spikes in individual Purkinje cells and their relationship between nearby cells remain elusive. In recent years, various genetically-encoded voltage indicators (GEVI), which can detect membrane potential changes, have been used to visualize spiking in individual neurons. In this study, we expressed a GEVI in Purkinje cells using adeno-associated virus (AAV) and examined whether simple spikes of Purkinje cells could be measured by voltage imaging in acute cerebellar slices. An AAV expressing a GEVI was pressure-injected into the cerebellum of wild-type C57BL/6J mice at P0. When the mice were 3 to 4 weeks old, acute sagittal cerebellar slices of 300 micrometers thick were prepared. Voltage imaging was performed under an upright microscope with a low noise sCMOS camera at a 200 – 400 Hz sampling rate. The membrane potential was recorded directly from the soma of Purkinje cells using the whole-cell recording technique, and the fluorescence changes in the soma were recorded simultaneously. The fluorescence change was highly correlated with the membrane potential change, with a steep rise and decay at the timing of the simple spike. Furthermore, the fluorescence change followed the subthreshold membrane potential change. These results showed that the simple spikes of Purkinje cells could be observed by voltage imaging with a GEVI. This method will allow us to simultaneously image simple spikes of multiple Purkinje cells and examine their spatiotemporal relationship in cerebellar slices and in vivo. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-005 蛍光寿命イメージングを用いたスパインに存在するホスファチジルイノシトール3,4,5-三リン酸(PIP3)の樹状突起から隔離の可視化 Spine PIP3 is sequestered from dendritic shafts *上田 善文(1)、小澤 岳昌(1) 1. 東京大学大学院理学系研究科 *Yoshibumi Ueda(1), Takeaki Ozawa(1) 1. The University of Tokyo Keyword: Fluorescence lifetime imaging, Synapse, Two-photon microscopy, Phosphatidylinositol 3,4,5-trisphosphate Phosphatidylinositol 3,4,5-trisphosphate (PIP3) regulates a broad spectrum of cellular functions as a lipid second messenger and is generated by phosphatidylinositol 3-kinase (PI3K) in response to hormones and neurotransmitters. For example, in neuronal cells, PI3K/PIP3 regulates neurite formation, the polarity of pyramidal neurons, dendritic arborization and axon growth . Additionally, spine PIP3 regulates synaptic function by maintaining α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor clustering. To exert these effects, the subcellular distribution of PIP3 is important because PIP3-binding proteins such as Akt, WASP family verprolin homologous protein and guanine nucleotide exchange factor small G proteins induce signaling by associating with PIP3. Dendritic spines are thought to be the primary sites of learning and memory in the brain. We previously developed a fluorescence lifetime-based PIP3 probe termed FLIMPA3 and demonstrated PIP3 enrichment within the dendritic shafts of CA1 neurons in hippocampal organotypic slices (Ueda Y et. al. J Neurosci 2013, 33:11040-11047.). However, the extent to which PIP3 compartmentalizes into dendritic spines in conditions in which PIP3 is increased remains unknown. Here, we investigated PIP3 dynamics in dendritic spines and shafts in response to glutamate stimulation and membrane depolarization and during neuronal development in hippocampal pyramidal neuronal cells. We found revealed that PIP3 accumulation in dendritic spines is strictly controlled even in an active state in which PIP3 is increased by glutamate stimulation and high potassium-induced membrane depolarization. Time-course PIP3 analysis clarified the gradual PIP3 accumulation in dendritic spines over days during neuronal development. Collectively, these results deepen our understanding of PIP3 dynamics in dendritic spines, and the dysregulation of this signal sequestration could cause neuronal diseases and mental disorders, such as autism spectrum disorder. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-006 成熟した初代神経細胞におけゲノム編集効率の発光を用いた定量的評価方法 Quantitative evaluation of genome editing efficiency by CRISPR/Cas9-mediated homologous recombination in mature primary neurons by dual luminescence. *清末 和之(1)、渡辺 布美(1) 1. 産業技術総合研究所 *Kazuyuki KIYOSUE(1), Fumi watanabe(1) 1. National Institute of Advanced Industrial Science and Technology (AIST) Keyword: genome edit, HDR, Cas9 Genome editing tools, especially the CRISPR/Cas9 system, have become powerful tools for investigating gene function in the field of neuroscience research through endogenous gene knockout and exogenous gene knockin. Although these tools have been successfully applied to fertilized eggs and early embryos, their application to post-mitotic cells, especially mature neurons, is still challenging due to the low efficiency of genome editing by homology-dependent recombination. We have previously reported homology-dependent gene incorporation into mature neurons by the CRISPR/Cas9 system and investigated ways to improve genome editing in mature neurons. As a result, we were able to evaluate the efficiency of genome integration by manually counting GFP fluorescently labeled neurons or by immunofluorescence detection using an integrated gene-specific antibody. This method was excellent for identifying whether genome-edited cells were neurons by morphological features, but it was time consuming. Therefore, to increase assay efficiency, we employed a dual-luciferase method using split nano-nanoluciferase. A donor gene (HiBit) encoding the N-terminal 13 amino acids of the nanoluciferase was used, and the expressed HiBit was combined with LgBit, a C-terminal split nanoluciferase, to form a functional nanoluciferase that was detected. To compensate for variations in the number of transfected reporter plasmids, the emission signal of the nanoluciferase was normalized by the emission of the co-transfected luciferase. We confirmed that this method worked well in neurons and attempted to identify factors that improve the efficiency of genome integration in mature neurons. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-007 コンピュータシミュレーションは樹状突起Ca(2+)シグナルがヘテロシナプス可塑性のために重要であることを示唆する Computer simulation suggests the importance of dendritic Ca(2+) signaling for heterosynaptic plasticity *浦久保 秀俊(1,2)、Laxmi Kumar Parajuli(4)、深澤 有吾(3)、岡部 繁男(4)、石井 信(2)、窪田 芳之(1) 1. 生理学研究所 脳機能計測・支援センター、2. 京都大学大学院情報学研究科、3. 福井大学医学部、4. 東京大学大学院医学系研究科 *Hidetoshi Urakubo(1,2), Laxmi Kumar Parajuli(4), Yugo Fukazawa(3), Shigeo Okabe(4), Shin Ishii(2), Yoshiyuki Kubota(1) 1. Support Cent Brain Res, Natl Inst Physiol Sci, Aichi, Japan, 2. Grad Sch Info, Kyoto U, Kyoto, Japan, 3. Fac Med Sci, U Fukui, Fukui, Japan, 4. Grad Sch Med, U Tokyo, Tokyo, Japan Keyword: synaptic plasticity, computer simulation, dendrite, Ca2+ signaling Synaptic plasticity, a neural basis of learning and memory, is induced by the activity of intracellular signaling molecules in a particular spatial domain: postsynaptic spine. Because the postsynaptic spines can hold the activity of signaling molecules, synaptic plasticity occurs in a stimulation-site specific manner (homosynaptic plasticity). However, this spatial hold is not perfect, and the stronger stimulations can further induce synaptic plasticity in neighboring unstimulated spines. This characteristic is called as heterosynaptic plasticity. For the induction of synaptic plasticity, Ca2+ signaling is known to be primarily important. However, no studies have addressed how the Ca2+ signaling is spread into the neighboring spines to induce heterosynaptic plasticity. We here try to address it by computer simulation, based on the shape data obtained from three dimensional volumetric images of electron microscopy. In the simulation, we first gave glutamate stimulation to activate N-methyl-D-aspartate receptors (NMDARs) at individual spines, and confirmed NMDAR-mediated Ca2+ influx. If the stimulated spines had endoplasmic reticula (ERs), we further observed Ca2+ spikes in those spines. This was because the glutamate activated the peri-synaptic signaling of mGluR→ Gq→ PLCβ → IP3, which triggered the mechanism of Ca2+-induced Ca2+ release (CICR) from inositol 1,4,5-trisphosphate receptors (IP3Rs) on the ERs. Then, we gave repetitive stimulation of glutamate to multiple spines simultaneously, which is known as a strong stimulation to induce heterosynaptic plasticity. This stimulation led to the overflow of IP3 from the stimulated spines into parent dendrites, resulting in a dendritic Ca2+ spike due to dendritic ERs. The spatial extent of the Ca2+ spike was determined by the spread of dendritic IP3, and the Ca2+ spike further activated a downstream molecule, calcineurin (CaN). CaN is known as a requirement of heterosynaptic plasticity, in particular, long-term depression (LTD). Indeed, the spatial extent of the CaN activity was compatible with that of experimentally-observed CaN as well as that of heterosynaptic LTD. These results suggest the importance of IP3 as a determinant of spatial rules in heterosynaptic plasticity. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-008 シナプス後膜肥厚のタンパク質群の液－液相分離による集合と離散 Segregation of postsynaptic proteins via liquid-liquid phase separation and their dissociation *細川 智永(1)、劉 品吾(1)、林 康紀(2)、木下 専(1) 1. 名古屋大学大学院理学研究科、2. 京都大学大学院医学研究科 *Tomohisa Hosokawa(1), Pin-Wu Liu(1), Yasunori Hayashi(2), Makoto Kinoshita(1) 1. Grad Sch Sci, Nagoya Univ, Nagoya, Japan, 2. Grad Sch Medicine, Kyoto University, Kyoto, Japan Keyword: Liquid-liquid phase separation, synapse, synaptic plasticity, postsynaptic density Recent progress on super resolution microscopy revealed that the distribution of postsynaptic proteins in PostSynaptic Density (PSD) is not homogenous but forming segregated clusters, called nanodomain, with the same protein molecules and their binding partners. Especially, the nanodomains of ionotropic glutamate receptor such as AMPAR and NMDAR are critical to understand the efficacy of synaptic transmission and synaptic plasticity. Previously we found that the activation and conformational change of Calcium/calmodulin dependent protein kinase II (CaMKII) by calcium/calmodulin results in the formation of protein condensate with NMDAR subunit GluN2B via liquid-liquid phase separation (LLPS). Also, autophosphorylation locks CaMKII in an active conformation, allows condensate to be remained after removal of calcium ion. On the other hand, PSD-95 forms protein condensate with both GluN2B and AMPAR auxiliary subunit Stargazin via LLPS. Activation of CaMKII in the presence of those proteins resulted in the segregation of GluN2B from PSD-95/Stargazin protein condensate. This would be the fundamental mechanism for the formation of nanodomains in PSD during excitatory stimulation. Considering bidirectional modulation of synaptic strength, we are also interested in the endogenous dissociation factor of PSD protein condensate. So far we identified Camk2n1/2 and Homer1a as the dissociation factor for PSD protein condensate. Camk2n1/2 dissociates CaMKII-driven protein condensate and Homer1a dissociates Homer-driven protein condensate. In this poster, we will introduce our previous study with recent results regarding the difference in the mechanism and significance of those endogenous dissociation factors. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-009 Regulation and underlying mechanisms of physiological synaptic plasticity by direct current modulation *Ching-Hsiang Chang(1), Chi-Wei Lee(1), Ming-Chia Chu(1), Chieh-Yu Chang(1), Tzu-Ning Peng(1), Tzu-Jung Yang(1), Hsiang Chi(1), Yen-Cheng Lin(1), Hui-Ching Lin(1,2) 1. Phy, NYCU, Taipei, Taiwan (R.O.C), 2. BRC, NYCU, Taipei, Taiwan (R.O.C) Keyword: Synaptic plasticity, N-methyl-D-aspartate receptors, in vivo Transcranial direct current stimulation, ex vivo direct current stimulation Synaptic plasticity, mediated by alternation of synaptic strength and connection, is the basis for neuronal function such as learning and memory in both health and disease conditions. Activation of excitatory glutamate receptors, N-methyl-D-aspartate receptors (NMDARs), was implicated in molecular pathways of plasticity change. Transcranial direct current stimulation (tDCS) is a non-invasive therapy widely progressed to improve several neuronal disorders by modification of synaptic plasticity in these decades. Besides, tDCS changes the resting membrane potentiation of local neurons by depolarization and hyperpolarization at the targeted area of the brain. To clarify whether tDCS treatment regulates synaptic plasticity in physiological condition, we applied in vivo tDCS model on C57BL/6 mice, and ex vivo DCS model on hippocampal brain slice. First, increased neuronal activation was detected in the hippocampus of C57BL/6 mice after in vivo tDCS treatment by immunofluorescence. Spatial working memory in C57BL/6 mice were modulated by in vivo tDCS treatment. Next, expressions of NMDAR and synaptic responses were upregulated in the hippocampus after ex vivo DCS treatment detected by western blot analysis and electrophysiological recordings, respectively. As mentioned above, in vivo and ex vivo data showed that DCS treatment performed excitatory effects on neuronal plasticity with involvement of NMDAR under physiological situations, indicating tDCS may be applicated on such diseases related to abnormal plasticity. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-010 Roadblock1 regulates FMRP function by promoting its degradation *Sara Emad El-Agamy(1), Laurent Guillaud(1), Keiko Kono(2), Yibo Wu(3,4), Marco Terenzio(1) 1. Molecular Neuroscience Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa, Japan., 2. Membranology Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa, Japan., 3. YCI Laboratory for Next-Generation Proteomics, RIKEN Center of Integrative Medical Sciences, Kanagawa, Japan., 4. Faculty of Science, University of Geneva Sciences II, Geneva, Switzerland. Keyword: Dynein complex, Neuronal survival, mRNA translation, Proteostasis Cytoplasmic dynein is the main eukaryotic retrograde molecular motor and plays a critical role in highly polarized cells such as neurons. Roadblock1 (DYNRLB1), one of three light chains of the dynein complex, was recently shown to be crucial for the survival of adult sensory neurons. However, the molecular mechanisms underlying its function remain elusive. We utilized a proximity-dependent biotinylation approach coupled with mass spectrometry to identify DYNLRB1 interactors in adult dorsal root ganglia (DRG). Among the candidates, the Fragile X mental retardation protein (FMRP), an RNA-binding protein linked to neurodevelopmental and neurodegenerative diseases, was selected for validation. We used proximity ligation to confirm the direct interaction between FMRP and DYNRLB1. To assess the impact of DYNLRB1 on FMRP dynamics and function, we combined shRNA-mediated silencing of DYNLRB1 with pharmacological treatments that target the proteasomal and lysosomal degradation pathways. Interestingly, DYNLRB1 knockdown significantly reduced FMRP proteostasis, leading to an increase in the number of FMRP granules in the somatic compartment of DRG neurons and repressing the translation of MAP1B, a microtubule binding protein and one of the primary mRNA targets of FMRP. Our findings suggest that DYNLRB1-FMRP interaction controls FMRP function through the promotion of its targeted degradation. This mechanism might play a significant role in the development or progression of neurodegenerative disorders. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-011 タバコ煙成分の一種であるメチルビニルケトンが神経細胞の発達と機能に与える影響 The Influence of Methyl Vinyl Ketone, a Component of Cigarette Smoke, on Neuronal Development and Function *周 至文(1)、菅原 大夢(1)、佐藤 蓮(1)、大西 拓(1)、小竹 皓貴(1)、松井 双葉(1)、乘本 裕明(1) 1. 北海道大学大学院医学研究院 *Zhiwen Zhou(1), Hiromu Sugawara(1), Ren Sato(1), Hiraku Onishi(1), Koki Kotake(1), Futaba Matsui(1), Hiroaki Norimoto(1) 1. Grad Sch Med, Hokkaido Univ, Hokkaido, Japan Keyword: axon, neuronal toxicity, memory, glia Smoking tobacco use is a major risk factor for cancers, vascular diseases, respiratory diseases, and neurological abnormalities and diseases. One of the toxic substances in tobacco smoke is methyl vinyl ketone (MVK), which is also released to the environment via industrial use and automobile exhaust. However, the effect of MVK on neurons has not been investigated. Here, we investigated the influence of MVK on neuronal development and functions both in vitro and in vivo. First, we determined whether MVK treatment negatively affects neuronal survival and axonal morphogenesis using primary hippocampal neuronal cultures. We observed that MVK caused neuronal death in a concentration-dependent manner. In the meantime, the results of morphology analysis suggests that low concentrations of MVK inhibit the branching of axons specifically and that high concentrations of MVK inhibit both the branching and outgrowth of axons. Then we investigated the influence of MVK on brain functions in vivo. Systematic treatment with MVK disrupted hippocampus-dependent spatial memory at a lower dose and reduced exploring behavior at a higher dose. Furthermore, in order to unveil the mechanism underlying the influence of MVK on memory, we used immunohistochemistry to check whether MVK affected neuronal survival, neuronal activity and glia functions in the hippocampus. In conclusion, we found that MVK is harmful and disruptive to neuronal development and functions. Even at a low concentration, MVK can negatively impact developing neurons. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-012 セプチン細胞骨格中核サブユニットSEPT7は前/後/傍シナプス膜ドメインに局在し、非筋型ミオシンIIB/MYH10と相互作用する The pivotal septin subunit SEPT7 localizes to pre-/post-/peri-synaptic membrane domains and interacts with MYH10/nonmuscle myosin IIB *木下 専(1)、三井 利来(1)、鈴木 絢子(1)、岩野 大渡(1)、深澤 有吾(2)、上田-石原 奈津実(1)、細川 智永(1) 1. 名古屋大学大学院理学研究科、2. 福井大学学術研究院医学系部門 *Makoto Kinoshita(1), Riku Mitsui(1), Ayako Suzuki(1), Daito Iwano(1), Yugo Fukazawa(2), Natsumi Ageta-Ishihara(1), Tomohisa Hosokawa(1) 1. Nagoya Univ Grad Sch Sci, 2. Fukui Univ Sch Med Sci Keyword: cytoskeleton, dendritic spine, presynaptic terminal, glial scaffold Plasma membranes of neurons and glia are segmented into nanodomains with diverse protein/lipid compositions and functions. The recruitment, confinement, and organization of membrane-associated proteins at each domain are regulated by the underlying network of cytoskeletal/scaffold proteins termed ‘membrane skeleton’. While submembranous actomyosin plays a key role in the membrane skeleton, the role of unconventional cytoskeletal filaments composed of oligomers of septin GTPases is poorly understood despite their implications in neuronal/glial functions and neuropsychiatric disorders. To address this, we conducted fine mapping of the core septin subunit SEPT7 in the dentate gyrus of mouse hippocampus by pre-embedding immunogold labeling method followed by serial section transmission electron microscopy (ssTEM) and 3D reconstruction. Immunogold signals for SEPT7 abound beneath plasma membranes of axons (shafts and terminals), dendrites (spine bases and necks), and glial processes. SEPT7 is composed of a globular GTPase domain and a flexible coiled-coil (CC) domain, both of which are known to contribute to the tight dimerization with SEPT6. We hypothesized that one of physiological phosphorylation sites in the CC (T426) could alter the biochemical property of SEPT7. As an unbiased screening of non-septin interacting partners of SEPT7, we conducted parallel affinity-purification of mouse brain lysate with four CC domains, wild type (WT), nonphosphorylable (T426A), and phosphomimetic (T426D and T426E) mutants, each fused to GST. Comparative LC-MS/MS analysis identified MYH10 as a major polypeptide that bound preferentially to WT and T426A . Given that MYH10 and SEPT7 abound in spine neck/base, and are co-immunoprecipitated from brain lysates, we tested tripartite interaction among CCs of SEPT6, SEPT7, and MYH10. Heterodimer of SEPT6CC and SEPT7CC pulled down MYH10CC, which was significantly diminished when SEPT7CC is replaced with the T426D mutant. These data indicate that the phosphorylation of SEPT7 at T426 weakens the interaction between SEPT7-containing septin oligomers and cortical actomyosin. We speculate that this mechanism facilitates membrane domain remodeling at spine neck/base. Currently, we are exploring contexts that cause T426 phosphorylation, responsible kinases, and physiological significance. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-013 BMP4欠損が前脳基底部からのコリン作動性投射に及ぼす影響 Effects of BMP4 defect to cortical cholinergic projections from the basal forebrain. *東 誉人(1)、岡部 繁男(2)、小林 靖(1) 1. 防衛医科大学校解剖学講座、2. 東京大学大学院医学系研究科 *Takahito Higashi(1), Shigeo Okabe(2), Yasushi Kobayashi(1) 1. Dept of Anatomy,Nat Def Med College, Saitama, Japan, 2. Grad Sch Med, Univ of Tokyo, Tokyo, Japan Keyword: CHOLINERGIC PROJECTION, BMP4, CHOLINERGIC NEURAL CIRCUIT The control of axonal trajectories is essential in the normal development of neural circuits. Axonal growth is induced by various factors: some of them act locally and others form concentration gradient to guide axons. However, precise mechanisms of axonal guidance still remain unclear. Bone morphogenetic proteins (BMPs) are members of the transforming growth factor (TGF)-b superfamily, and involved in establishing the basic body plan. Secreted BMP bind to their receptors and activate a canonical or non-canonical signaling pathway. Recently, it is shown that BMP signaling molecules are involved in CNS patterning, axon pathfinding and synapse remodeling in mammals. We also reported that BMP4 has an important role in selective synapse destabilization and elimination of axons with an autocrine secretion manner. Although we revealed that BMP4 functions in the early phase of synapse development, it remains elusive whether BMP4 regulates the axonal projection. Here, we show that BMP4 can restrict the neocortical target area of cholinergic projections from the basal forebrain in mice. Cholinergic neurons in the basal forebrain innervate the entire extent of the neocortex. We therefore hypothesized that a long-range cholinergic projection is regulated by BMP4. To verify our hypothesis, we generated BMP4 conditional knockout mice (BMP4 cKO) using cholinergic neuron Cre driver: ChAT (Choline Acetyltransferase)-Cre. These mice were born in the ratio by Mendelian genetics and they exhibited normal body size in adulthood. However, the forebrain length in BMP4 cKO mice was slightly longer than in BMP4 floxed control mice. We next investigated the number of cholinergic axonal boutons in the neocortex using vesicular acetylcholine transporter (VAChT) immunohistochemistry. Interestingly, BMP4 cKO mice exhibited markedly increased VAChT positive boutons in the neocortex, especially in the prefrontal cortex. Similar terminal increase was also observed in the striatum, amygdala and hippocampus. To further evaluate whether these changes were caused by the increase of cholinergic neurons, or of their axonal branching, we currently investigated the numbers and axonal morphology of cholinergic neurons in the basal forebrain. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-014 細胞放出分子の新規可視化技術：ヒト培養真皮線維芽細胞のコラーゲン放出とUV-A照射の影響 Novel technology for secreting molecular observation: collagen release and effects of UV-A irradiation on human cultured dermal fibroblasts *佐竹 繁寿(1)、新谷 太地(1)、仲田 しずか(1)、小林 和人(2)、穂積　 直裕(1)、吉田 祥子(1) 1. 豊橋技術科学大学 、2. 本多電子株式会社 *Shigehisa Satake(1), Taichi Shintani(1), Shizuka Nakada(1), Kazuto Kobayashi(2), Naohiro Hozumi(1), Sachiko Yoshida(1) 1. Toyohashi University of Technology , 2. Honda Electronics Co., Ltd. Keyword: Ultrasonic microscopy, Fibroblast, Collagen, UV-A Ultrasonic microscopy (UM) enables the noninvasive measurement of living tissue and cells. Previous studies have shown changes in intracellular structure, especially cytoskeleton, mitochondria, and nucleus. Recently, using UM, we have visualized the distribution of released collagen from living cultured fibroblast cells and observed an increase in collagen release treated with ascorbate. Because the acoustic impedance values (AIM) acquired with the UM reflect the density and viscoelasticity of materials, collagen or other extracellular matrix molecules are able to visualize with UM. AIM of the extracellular area was increased populationally with the collagen quantification in the culture medium by ELISA. Fibroblasts in the dermis of the skin play an important role in the skin's aging process by secreting collagen and other extracellular matrices. In this study, we investigated the effect of UV-A irradiation on human cultured fibroblasts using UM. UV-A is ultraviolet radiation (UVR) with 320-400 nm wavelength. 90% of the UVR we are exposed to daily is UV-A, and UV-A causes a decrease in skin color and firmness, skin roughness, and skin irritation. We irradiated cultured fibroblast with UV-A at an irradiation intensity of 420 μW/cm2 on day 5 in vitro (DIV 5) for 0 (control), 5, 10, 20, 30, and 40 min at once. UV-A irradiation decreased fibroblast viability significantly compared to the control group, whereas the amount of collagen per cell in the culture medium was little decreased in the control group. The UM observations showed a significant increase in AIM in UV-A-irradiated cell nuclei, cytoskeleton, and extracellular regions; intracellular organelle would change their conformation to oppose UV irradiation. These suggested that the UM is useful for visualizing the secreting molecules and intracellular conditions of living cells under UV-A irradiation. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-015 Plagl2およびDyrk1a遺伝子の発現制御による老化神経幹細胞の機能的若返り Functional rejuvenation of aged neural stem cells by Plagl2 and anti-Dyrk1a activity *貝瀬 峻(1,2,3)、福井 雅弘(1,2,3)、末田 梨沙(2,4,6)、朴 文惠(2,4)、山田 真弓(2,4,5)、小林 妙子(2,3,4)、今吉 格(2,4)、影山 龍一郎(1,2,3,4,5) 1. 理化学研究所 CBS、2. 京都大学ウイルス再生医科学研究所、3. 京都大学大学院医学研究科、4. 京都大学大学院生命科学研究科、5. 京都大学物質-細胞統合システム拠点、6. セインズベリー・ウェルカムセンター *Takashi Kaise(1,2,3), Masahiro Fukui(1,2,3), Risa Sueda(2,4,6), Wenhui Piao(2,4), Mayumi Yamada(2,4,5), Taeko Kobayashi(2,3,4), Itaru Imayoshi(2,4), Ryoichiro Kageyama(1,2,3,4,5) 1. RIKEN CBS, Saitama, Japan, 2. Institute for Frontier Life and Medical Sciences, Kyoto University, Kyoto, Japan, 3. Graduate School of Medicine, Kyoto University, Kyoto, Japan, 4. Graduate School of Biostudies, Kyoto Univerisy, Kyoto, Japan, 5. Institute for Integrated Cell-Material Sciences, Kyoto University, Kyoto, Japan, 6. Sainsbury Wellcome Centre, University College London, London, UK Keyword: adult neurogenesis, aged brain, neural stem cell, Plagl2 In the hippocampus of the aged mouse brain, not only the number of neural stem cells (NSCs) but also their activation rate and/or neurogenic potential significantly decline, leading to cognitive dysfunctions. This decline involves up-regulation of senescence-associated genes, but inactivation of such genes failed to reverse aging of hippocampal NSCs. Because many genes are up-regulated or down-regulated during aging, manipulation of single genes would be insufficient to reverse aging. Here we searched for a gene combination that can rejuvenate NSCs in the aged mouse brain from nuclear factors differentially expressed between embryonic and adult NSCs and their modulators. We found that a combination of inducing the zinc finger transcription factor gene Plagl2 and inhibiting Dyrk1a, a gene associated with Down syndrome (a genetic disorder known to accelerate aging), rejuvenated aged hippocampal NSCs, which already lost proliferative and neurogenic potential. Such rejuvenated NSCs proliferated and produced new neurons continuously at the level observed in juvenile hippocampi, leading to improved hippocampus-dependent learning and cognition. Live-imaging analysis showed that quiescent NSCs start to express proneural gene Ascl1 in oscillatory manner, a hallmark feature of active NSCs, by inducing Plagl2 and inhibiting Dyrk1a. Epigenome and transcriptome analyses indicated that this gene combination induces up-regulation of embryo-associated genes and down-regulation of age-associated genes by changing their chromatin accessibility, thereby rejuvenating aged dormant NSCs to function like juvenile active NSCs. Thus, aging of NSCs can be reversed to induce functional neurogenesis continuously, offering a way to treat age-related neurological disorders. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-016 DCGANによるCG昆虫写真の生成 Generation of CG images of insects by DCGAN *前野 杏美(1)、三浦 佳二(1) 1. 関西学院大学 *Ambi Maeno(1), Keiji Miura(1) 1. Kwansei Gakuin University Keyword: deep learning, DCGAN, ants, insects In recent years, animal images are often available from various databases on the Internet. Although the databases for free materials for using in slides etc. have been enriched, it is still difficult to obtain the image realistically when, for example, we want an image of a particular species of ant taken from a certain orientation. In particular, ants have a small body length, so it is difficult to take photos. Therefore, the number of images of ants is extremely small. In this study, we aimed to construct GAN that can artificially generate the images of the ants that satisfy various requests on attributes such as species and orientations. Before generating CG images of ants, we used the Fashion-MNIST dataset to confirm that DCGAN is more accurate than GAN. Initially, we plotted Fashion-MNIST dataset and built generator and discriminator. Next, we trained 50 epochs on each. After the training was completed, we had one CG image generated. The results showed that DCGAN was more accurate than GAN, so we decided to use DCGAN to generate the images. To generate CG images of ants, initially we collected various types of images of ants and resized all the images (a total of 717 images) to 32x32 by Python. Next, we built the discriminator and generator. After that, we trained 900 epochs on each of the resized images. After the training was completed, we had one CG image generated. As a result, we were able to generate CG images that could be mistaken for real ants using DCGAN. However, the accuracy was lower than images trained and generated using Fashion-MNIST dataset and there were also scattered images that were unrecognizable as ants. We also found it difficult to directly control the attributes of the images with the plain DCGAN. Therefore, in the future, we would like to make it possible to control the attributes and collect more training images in order to create highly accurate and diverse CG photos of ants. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-017 非接触ミリ波センサを用いた感情認識システムの開発 Development of emtion recognition system using a contactless millimeter-wave sensor *宮崎 淳吾(1)、髙野 楓子(1)、眞野 博彰(1)、小林 賢也(1)、西井 裕亮(1)、荒川 智哉(1)、黒田 淳(1)、熊田 孝恒(2) 1. 京セラ株式会社 先進技術研究所、2. 京都大学情報学研究科 *Jungo Miyazaki(1), Fuuko Takano(1), Hiroaki Mano(1), Masaya Kobayashi(1), Yusuke Nishii(1), Tomoya Arakawa(1), Jun Kuroda(1), Takatsune Kumada(2) 1. Advanced Technology Research Institute, Kyocera Corporation, 2. Graduate School of Informatics, Kyoto University Keyword: emotion, affect, millimeter-wave sensor, artificial intelligence With the spread of COVID-19, people have been forced to wear masks and practice physical distancing in public places. Under these circumstances, it may be difficult for people to estimate the emotional states of others based on limited facial expressions and other external features. Although tools for online communication have become widespread, it is similarly difficult to estimate the internal states of others because visual information displayed on electronic devices is limited, which substantially reduces the quality of communication. To address this issue, methods for estimating human emotions from internal characteristics such as physiological signals have been investigated. However, most of these methods require sensors to be attached to the body of the person and carry the risk of infection. Therefore, we are developing a system to estimate a person’s internal states in a contactless manner. This system includes an originally developed compact millimeter-wave sensor. This sensor can be placed on a desk and detect heart sounds, enabling us to easily measure RR intervals widely used in heart rate variability analysis for affective computing and medicine. We investigated the statistical properties of the heart rate data measured by our millimeter-wave sensor and compared it to that of a contact electrocardiography sensor. In addition, we conducted an experiment to elicit emotional states that are important in online communication, such as engagement. During this experiment, we measured RR intervals with the millimeter-wave sensor. We estimated the emotional states from the RR intervals, and confirmed that it outperforms the baseline conventional method. We hope that the system can be incorporated into online communication platforms in order to enrich people’s communication. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-018 CIFAR10の画像分類率を最大化する深層ネットワーク構造の探索 Exploring deep network structures that maximize classification performance of CIFAR10 images *株本 あやめ(1)、三浦 佳二(1) 1. 関西学院大学理工学研究科 *Ayame Kaumoto(1), Miura Keiji(1) 1. Grad Sch Science and Engineering,KwanseiGakuinUniversity,Hyogo,Japan Keyword: deep learning, CNN, CIFAR10, visual cortex Although image classification technology using deep learning is advanced, a classification accuracy gap with humans exists. In order to compare the image classification performance with humans, we used Convolutional Neural Networks, which mimic the neuronal connections in the human visual cortex, and attempted to form a network model on TensorFlow that is close to the human image classification accuracy. Especially, we aimed at exploring the deep network structures that maximize the classification performance of CIFAR10 images. We attained the perfect training performance (100%) when we layered three visual cortices. The generalization performance also increased to 82.03% by having the “tertiary” visual cortex. We repeated the test with additional layers for the fourth visual cortex and beyond, but could not confirm any increase in generalization performance due to overlearning. As we considered that over-learning prevented an increase in generalization performance, we set up Dropout and Batch Normalization in each layer and added a weight decay to adjust the weights in the parameters. Accuracy increased by 4.4% from the previous experiment, with a maximum accuracy value of 86.43%. To visualize the classification status at each layer, the principal component analysis was performed to compare all 27 layers, and we found that the neural representation of labels was insufficient in the layers in the secondary visual cortex, but it was well-separated in the layers in the tertiary visual cortex. Based on these results, we considered that image classification with CNNs requires at least the third-order visual cortex as far as the image size is moderate (32x32). 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-019 Echo state networkに基づく時系列予測モデルによる暗黙的知覚学習 An echo state network-based time-series prediction model for implicit learning *後藤 優仁(1)、北城 圭一(2) 1. 総合研究大学院大学、2. 生理学研究所 *Yujin Goto(1), Keiichi Kitajo(2) 1. The Graduate University for Advanced Studies (SOKENDAI), 2. National Institute for Physiological Sciences Keyword: Reservoir computing, Implicit learning, auditory perception, predictive coding We propose an echo state network-based time-course prediction model for the mechanism of implicit sensory learning. A recent human study using an unsupervised implicit learning paradigm reported that the human brain could quickly learn auditory random noise stimuli (Agus, et al. 2010). In this study, participants were presented with 1000ms auditory noise. The stimulus set contains four types of noise stimuli, which can be defined by two features. The half of the stimuli set have seamless noise segment repetition (Repeated noise; RN), the first 500ms of the stimuli are the same as the latter, and the other half have no repetition (Noise; N). Specific patterns of RN and N occasionally reoccurred through the experiment (Referenced; Ref), and other stimuli were original. The participants were asked to detect the repetition (RN) of stimulus. Even though the participants were unaware of the existence of RefRN, they showed selective higher detection scores for the RefRN than RN stimuli. In the context of predictive coding, humans perceive the world based on predictive information rather than actual sensory information. Therefore, in the above experiment, it is speculated that learning increases prediction accuracy for RefRN, leading to consistent brain activity across distinct trials and high task performance. We investigated how the prediction accuracy for noise time series data could be improved by learning. For the computational model, we used ESN. The output joint weight matrix was trained to predict one time-step ahead time courses of input noise signals as output. During training, RefRN was input multiple times and other signals, RN and N, were input only once as training data. After learning, each input stimulus was re-input as a test stimulus, and the root mean squared error (RMSE) between the output predicted value and the correct data was calculated. The RMSE was smaller for the RefRN and larger for the remaining. The ESN-based model exhibited increased accuracy of future prediction for RefRN. When the input stimuli contained temporal repetitions, the prediction signal also showed repetitive patterns. This means that learning is associated with consistency in responses to the identical input stimulus, overcoming the influence of initial conditions of the system. Also, the increased consistency of model responses when an identical input reoccurred enables accurate future prediction as well as high detection performance. The results suggest that the prediction accuracy for noisy sensory inputs could be improved by implicit learning based on the dynamical features of the brain. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-020 神経クラスタ推定法の高次元への一般化 Generalization of neuronal ensemble inference method to higher dimensions *木村 俊(1)、竹田 晃人(1) 1. 茨城大学大学院理工学研究科 *Shun Kimura(1), Koujin Takeda(1) 1. Graduate School of Science and Engineering, Ibaraki Univ. Keyword: Selberg type integral, Bayesian inference, functional neuronal ensemble It is widely known that various brain functions essential to life activities are realized by the interaction of innumerable neurons. To reveal the mechanism of these brain functions, the neuronal network structure or functional connectivity strength among neurons has been investigated in wide field of neuroscience. On the other hand, it has been suggested that functional neuronal ensembles, which have influential hub neurons in their centers, are involved in efficient information processing among neurons. Therefore, for the mechanisms of brain functions, it is necessary to infer ensembles and intra/inter-network structure of theses ensembles. Specifically, in the field of data analysis, the inference method for ensembles or connectivity strength from calcium imaging fluorescence data is proposed to meet the above-mentioned demand. For ensemble structure, the method based on synchronization of neuronal activity is known. By this method, the domain of strong connectivity or dense network structure in the network can be inferred under the assumption that neurons having similar function will show synchronized activity. More precisely, the ensemble structure is inferred by a Bayesian method based on generative model reflecting prior knowledge on activity synchronization. As a feature of this method, a stochastic process called Dirichlet process allows us to infer the number of ensembles, ensemble structure, and time-series activity of each ensemble from input time-series neuroactivity data without knowing the number of ensembles in advance. However, in this generative model, dynamical activity of each ensemble is inferred by temporal average of activity in the ensemble. This leads to the problem of unstable inference accuracy for the data showing rapid change of activity frequency in the measurement period. In this presentation, we propose a generative model with improved expressibility for resolving the above-mentioned problem. Specifically, in our generative model, integrand in Euler's beta integral is replaced by the one in Selberg type integral, by which activity frequency and synchronization can be modeled independently at each measurement time. Selberg type integral is the generalization of Euler's beta integral to multiple dimensions, whose application is known in random matrices. We will validate our generative model with Selberg type integral by applying the original/proposed methods to multiple data sets having different activity features. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-021 Explaining Biological Neural Dynamics with Continuous Time Recurrent Neural Networks Trained on Behavioral Tasks *Arthur Pellegrino(1), Angus Chadwick(1) 1. University of Edinburgh Keyword: Recurrent Neural Networks, Dynamical Systems, Control Theory, Low-Dimensional Manifolds Recurrent neural networks (RNNs) are used extensively throughout neuroscience, both by fitting them to neural recordings to infer latent dynamics, and by training them on behavioral tasks as a means of generating hypotheses for the solutions used by neural circuits. However, a link between these two approaches has yet to be drawn. In particular, methodologies for systematic comparison of the dynamics of recurrent neural networks trained on behavioral tasks to the neural responses of brain regions relevant to these tasks are currently lacking. The challenge of linking artificial and biological neural network dynamics is compounded by the substantial trial-to-trial variability in neural recordings that is not explained by task-relevant variables. This can be caused by variability in inputs to the region recorded, or more generally imperfect trial repetitions by the animal performing the task. However, such variability does not naturally arise in task-optimized recurrent neural networks, which obfuscates direct comparison of biological and artificial network dynamics on a per-trial basis. In the present work, we develop a method to compare RNNs dynamics to neural population activity on a trial-by-trial basis. This method consists of simultaneously optimizing an input control and a nonlinear warping function of the differential equation characterizing the RNN, on each trial, to maximize the variance of the neural data explained by the RNN dynamics on a shared low dimensional manifold. We apply our method to a set of continuous-time RNNs with distinct architectures, trained on the visual decision making task of the International Brain Laboratory. First, to validate our approach, we test whether the controls and warpings can account for variability happening within a model by applying the method to RNNs against a sample of their own dynamics during the task. Then, the different dynamic solutions learned by each architecture are compared to the neural recording of an animal performing the task. We thus illustrate how our work allows for the comparison of task-optimized recurrent neural networks to biological neural data at the level of single trials, enabling the generation and validation of hypotheses regarding the latent neural dynamics that underpin cognitive tasks. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-022 カーネル法による予算付き追加学習法とキノコ体出力ニューロンの学習メカニズムとの関係 A kernelized incremental learning on a budget and its connection with the learning mechanism of mushroom body output neurons *山内 康一郎(1)、平手 貴大(1) 1. 中部大学 *Koichiro Yamauchi(1), Takahiro Hirate(1) 1. chubu university Keyword: Incremental Learning on a budget, Kernel machine, mushroom body output neuron (MBON), forgetting and adaptation In this presentation, we try to discuss the a kernelized incremental learning method on a budget as a model of mushroom body output neuron of drosophila. In our previous work, we have developed a kernelized continuous learning on a budget (Yamauchi 2014, 2021). This was a pure engineering model that can be performed with limited memory space and low computing power such as built-in microcomputers. As the network size is limited because of its memory capacity, it cannot continue to save new samples forever. Instead, it must forget part of the ineffective memory to make space to adapt to an important new event. To that end, the core learning system must find a balance between forgetfulness and adaptation. Consequently, our proposed method achieves highly effective incremental learning under a limited memory capacity. 　It is well known that novel odor elicits strong activity of the mushroom body output neuron (MBON)-α'3. However, the activity is rapidly suppressed by the repeated exposure to the same odor (Hattori et al., 2017). This strongly suggests that the fruit fly conducts incremental learning of new odours. According to the anatomical point of view, drosophila seems to be able to memorize 20 odors (Hattori et al., 2017). We are very interested in how to manage the 20 olfactory capacity throughout the lifetime. In this presentation, we compare the behaviors of our proposed learning method and the MBON-α'3 against the learning task for new novel instances and discuss the incremental learning strategy of drosophila. References: D.Hattori, Y.Aso, Kurtis J. Swartz, Gerald M. Rubin, L.F. Abbott, Richard Axel . "Representations of Novelty and Familiarity in a Mushroom Body Compartment", Cell, vol.169, pp. 956--969, May, (2017). K.Yamauchi . "Incremental learning on a budget and its application to quick maximum power point tracking of photovoltaic systems", Journal of Advanced Computational Intelligence and Intelligent Informatics, vol.18, No.4, pp. 682--696, (2014). K.Yamauchi . "Quick Continual kernel learning on bounded memory space based on balancing between adaptation and forgetting", Evolving Systems, under review, September, (2021). 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-023 スパイキングニューラルネットワークを用いる強化学習のためのTD誤差表現の提案 A new formulation of temporal difference error in reinforcement learning for spiking neural networks. *吉村 英幸(1,2)、銅谷 賢治(2)、山﨑 匡(1) 1. 電気通信大学大学院情報理工学研究科、2. 沖縄科学技術大学院大学神経計算ユニット *Hideyuki Yoshimura(1,2), Kenji Doya(2), Tadashi Yamazaki(1) 1. Graduate School of Informatics and Engineering, The University of Electro-Communications, Tokyo, Japan, 2. Neural Computation Unit, Okinawa Institute of Science and Technology Graduate School, Okinawa, Japan Keyword: Reinforcement Learning, Spiking Neural Network Reinforcement learning is a powerful machine learning algorithm. The reinforcement learning framework consists of an environment and an agent. The agent takes an action depending on a current state of the environment, which in turn delivers an updated state and a reward to the agent. The objective of reinforcement learning is to have the agent learn the optimal action sequence that maximizes rewards obtained from the environment through trial and error. In recent years, deep reinforcement learning, which combines reinforcement learning and deep neural networks, has made remarkable progress, such as surpassing human performance in the game of Go and various other domains. On the other hand, reinforcement learning using spiking neural networks has not been sufficiently studied. Several existing methods achieved to implement reinforcement learning on spiking neural networks, but certain architecture using purely spike-based learning have not been established. Developing a purely spike-based learning method for reinforcement learning will be useful, because once we develop such a method, we can implement it on neuromorphic hardware expected for low energy consumption. In this study, we focused on how to implement temporal difference error, the core learning signal of reinforcement learning, as a spiking neural network. First, we proposed a new formulation of temporal difference error; divisive-TD. This formulation is obtained by a simple transformation from the conventional formula of temporal difference error. It has characteristics consistent with conventional temporal difference, and can be implemented as spiking neural networks by using pre synaptic short-term plasticity. Then, we showed an example of spiking neural networks for reinforcement learning using the divisive-TD. The network realizes an actor-critic method, and consists of three layers, the value layer, the TD layer, and the action layer. These layers are composed of only spiking neurons, and synaptic weights are learned by spike timing-dependent synaptic plasticity. Finally, we demonstrated that our method can solve basic reinforcement learning tasks with numerical simulation. In summary, our method successfully achieves reinforcement learning with simple spiking neural networks. This result does not only contribute to the development of machine learning using spiking neural networks, but also suggests that reinforcement learning can take place in various parts of the brain. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-024 A computational model of the claustrum in mice *Razvan Gamanut(1,2), Carlos Enrique Gutierrez(1,3), Kenji Doya(1) 1. Neural Computation Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa, Japan, 2. Department of Physiology, Monash University, Melbourne, Australia, 3. SoftBank, Tokyo, Japan Keyword: claustrum, cortex, computational model The claustrum is a group of neurons located underneath lateral temporal and caudal orbitofrontal cortex, and remains one of the least understood parts of the mammalian nervous system. One of the reasons is the complex geometry of the claustrum (a folded, thin layer of neurons, sandwiched between other cellular groups and white matter tracts), which creates specific challenges for experimentation. However, in recent years the claustrum has been studied intensely in mice, revealing many details about its cellular composition and dynamics, but still without a satisfactory mechanistic explanation of its function. This work investigates through computational simulations the internal mechanisms of claustrum function. To this end, we built a detailed in-silico model of the mouse claustrum. Specifically, we used NEST (nest-simulator.org) to create a network of spiking neurons, with parameters collected from biological data. The parameters for individual neurons include membrane capacitance and resistance, resting potential, action potential threshold, accommodation index, postsynaptic decays and weights of connections (Kim et al, 2016; White and Mathur 2018; Graf et al, 2020). As a suitable neuronal model which accommodates all of these parameters, we use the adaptive exponential integrate and fire (Brette and Gerstner 2005). We also connect pairs of neurons with frequencies depending on the physical distance between the neurons, as was found with paired whole-cell recordings in slices (Kim et al, 2016). We found that extrapolating these rules of connectivity in 3D leads to a heterogeneous distribution of inputs and outputs for each neuron, with a spatial bias for the numbers of connections consistent with experimental observations. These results shed light on the impact of geometrical and organisational constraints on the structure and function of claustrum. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-025 レザバー計算による音周波数弁別タスク時のマウス行動モデルの構築 Modeling mouse behavior during sound frequency discrimination task by reservoir computing *上岡 雄太郎(1,2)、Shuo Wang(1)、石津 光太郎(1)、船水 章大(1) 1. 東京大学 定量生命科学研究所、2. 8thCAL株式会社 *Yutaro Ueoka(1,2), Shuo Wang(1), Kotaro Ishizu(1), Akihiro Funamizu(1) 1. IQB, Univ of Tokyo, Tokyo, Japan, 2. 8thCAL Inc. Keyword: Behavior, Modeling, Reservoir computing How do animals adapt their behavior to survive in a constantly changing environment? One approach to answer this question is to model the brain that generates the movement of animals. In a decision-making paradigm, recent studies have found that brain activity is highly influenced by whole-body movement. Therefore, we used reservoir computing to model the brain that drive whole-body movements during a decision-making task. Reservoir computing is a type of Recurrent Neural Network that allows real-time inputs and outputs. It requires updating output weights but not weights of internal recurrent network. A tone-frequency discrimination tasks were performed to water-restricted mice. Mice chose the left or right spout after listening sound stimuli compromised of various frequencies. Mice were rewarded with water when they lick the right spout if the presented sound stimulus is high-frequency, or the left spout if low-frequency. During the tone-frequency discrimination task, the whole-body movement of the mouse was captured by four cameras located in front, back, left and right side of mouse. First, the body movements in these videos were analyzed with Deeplabcut (a deep learning-based body tracking software) and extracted the trajectory of 68 major body coordinates. We performed principal component analysis on the 68 components and found that 20 principal components (PCs) explain approximately 90% of all the trajectories. We trained a reservoir computing to predict the 20 PCs from the environmental inputs of sounds, spout movements, and rewards. Our preliminary result showed that the reservoir computing succeeded to model the average trajectory of mouse body movement. However, the model did not follow the trial-by-trial change of body movement. Further analyses are needed to construct a brain-like network model generating behavior. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-026 脳応答予測モデルを介して深層ニューラルネットの内部情報表現を脳様表現に変換する Transforming information representations of deep neural networks to brain-like representations via models for brain-response prediction *川畑 輝一(1,2)、王 佳新(1,2)、Blanc Antoine(1)、西本 伸志(1,2)、西田 知史(1,2) 1. 情報通信研究機構（NICT）未来ICT研究所 脳情報通信融合研究センター（CiNet）、2. 大阪大学 *Kiichi Kawahata(1,2), Jiaxin Wang(1,2), Antoine Blanc(1), Shinji Nishimoto(1,2), Satoshi Nishida(1,2) 1. Center for Information and Neural Networks (CiNet), Advanced ICT Research Institute, National Institute of Information and Communications Technology (NICT), 2. Osaka University Keyword: Deep learning, Humanness, Brain representation, Neuroimaging Deep neural networks (DNNs) have recently achieved sophisticated performance, in pattern recognition tasks, comparable with or, in some cases, higher than human performance. However, information representations of such DNNs still have non-negligible differences from those of the human brain with regard to, for example, visual object processing (Raman & Hosoya, 2020; Xu & Vaziri-Pashkam, 2021). This may cause DNN’s behavioral characteristics in pattern recognition that are distinctive from human ones (Goodfellow et al., 2015; Geirhos et al., 2018). Therefore, to improve the humanness of DNN’s recognition behaviors, which is an important factor for realizing human-centered artificial intelligence, we should make DNN’s information representations closer to brain ones. We address this issue by introducing a method for transforming DNN’s internal representations to brain-like ones via models for brain-response prediction. In our method, computational models are constructed to predict brain response to arbitrary audiovisual or linguistic inputs from the combination of DNN’s activation patterns to the inputs and the history of preceding brain response. Importantly, brain measurement with functional MRI for only several hours is used for the model construction, but afterwards no additional measurement is needed. To validate this method, we performed a variety of audiovisual and linguistic pattern recognition tasks using DNN’s representations transformed into brain-response representations via the models. We then compared the recognition patterns with when the representations were not transformed, in terms of the similarity with the recognition patterns directly using measured brain response (i.e., brain decoding); if the representations are closer to brain ones, the recognition patterns should be more similar to those of brain decoding. We found that the similarity with brain decoding became higher when DNN’s representations were transformed than when they were not transformed in both the audiovisual and linguistic tasks. This result indicates that the transformation into brain-response representations successfully makes DNN’s representations closer to brain ones independently of input modality. Our method has a great potential to improve the humanness of DNN’s behavior in various types of pattern recognition. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-027 Brain-like combination of feedforward and recurrent network components achieves prototype extraction and robust pattern recognition *Naresh Balaji Ravichandran(1), Anders Lansner(1,2), Pawel Herman(1) 1. KTH Royal Institute of Technology, 2. Stockholm University Keyword: Attractor network, Unsupervised Hebbian learning, Associative memory, Brain-like computing Associative memory has been a prominent candidate for the computation performed by the massively recurrent neocortical networks. Attractor networks implementing associative memory have offered mechanistic explanation for many cognitive phenomena. However, attractor memory models are typically trained using orthogonal or random patterns to avoid interference between memories, which makes them unfeasible for naturally occurring complex correlated stimuli like images. We approach this problem by combining a recurrent attractor network with a feedforward network that learns distributed representations using an unsupervised Hebbian-Bayesian learning rule. The resulting network model incorporates many neurocomputational properties: unsupervised learning, Hebbian plasticity, sparse distributed activations, sparse connectivity, columnar and laminar cortical architecture, etc. We evaluate the synergistic effects of the feedforward and recurrent network components in complex pattern recognition tasks on the MNIST handwritten digits dataset. We demonstrate that the recurrent attractor component implements associative memory when trained on the feedforward-driven internal (hidden) representations. The associative memory is also shown to perform prototype extraction from the training data and make the representations robust to severely distorted input. We argue that several aspects of the proposed integration of feedforward and recurrent computations are particularly attractive from a machine learning perspective. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-028 脳波には、正弦波に加えて等電位ゆらぎが含まれる：τとバースト Brain waves would include equipotential fluctuations in addition to oscillations as sine waves: τand burst *小山 千佳(1) 1. 酪農学園大学 *Chika Koyama(1) 1. Rakuno Gakuen University Keyword: New brain information reading technology, EEG, anesthesia, LFP One of the remaining issues in brain research will be the development of high-precision brain information reading technology. I recently developed new brain-wave indices, a sub-threshold wave“τ” and an above-threshold wave “burst”. A number of τ(Nτ) was maximum at a mean duration of τ(Mτ) of 2.5-3 sampling intervals. Its threshold (ThNτmax: microwave amplitude)and amplitude of burst (Abst) were correlated with the vigilance state. As there was no difference in the ratio of total τ at ThNτmax(about 30%) on 125Hz-1000Hz LFPs (Koyama et al, 44thJNSS), those τs fluctuate every few sample intervals even when the sampling frequency is increased. Therefore, τs at ThNτmax are presumed to be momentary equipotential states. In this study, I showed that the new indices were also useful for EEG by depth of anesthesia in dogs as well as for LFP in mice. Changes in new indices in increasing anesthetic dose, whose effect on neurons involves hyperpolarization, were examined and the principle of τwas explored. Methods I analyzed EEG sampled at 256Hz in 6 senior dogs and 6 young-adult dogs anesthetized with sevoflurane (SEV) at 2.0, 2.5, 3.0, 3.5, 4.0, 5.0 (%). For three 64s EEG data at each SEV, the number of τ(Nτ), mean duration of τ(Mτ) and mean amplitude of burst (Abst) at threshold every 0.5µV from 0.5µV to 10µV were computed. Data were analyzed using C and R programs with self-made scripts. Results With increasing threshold, Nτ increased and then decreased, and Abst decreased and then increased. Nτwas maximum at Mτof 10.3±2.1 (ms), similar to those in LFP. As SEV increased except BS>50% level, at the threshold where Mτ was near 2.5sample intervals (9.8ms), Nτdecreased (r=-0.84±0.09 and -0.93±0.05 (seniors and young-adults)) and Abst increased (r=0.85±0.04 and 0.94±0.05). As SEV increased from 2.0% to 5.0%, that threshold decreased (r= -0.93±0.04.8, -0.75±0.14).Sleep-like activity under light to moderate anesthesia is based on a bistable up-down pattern in cortical neurons that hyperpolarization and depolarization alternate. Burst suppression activity under deep anesthesia is based on a unimodal state that which hyperpolarization is predominant. Therefore, decreases of ThNτmax depending on anesthetics dose indicate that microwave amplitude is correlated with the uncertainty of membrane potentials. I suggest that by aligning Mt of 2-3 sampling intervals, τand burst would be used as indicators of equipotential fluctuation and active potential. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-029 自然動画提示に対する誘発脳波活動と深層ニューラルネットワーク応答の対応付け Analyzing a correspondence between human EEG and deep neural network responses to naturalistic video presentations *倉重 宏樹(1)、金子 順(1) 1. 東海大学情報通信学部 *Hiroki Kurashige(1), Jun Kaneko(1) 1. Sch Info Telecom Eng, Tokai University, Tokyo, Japan Keyword: ELECTROENCEPHALOGRAM (EEG), DEEP NEURAL NETWORKS, NATURALISTIC VIDEO, HUMAN Deep neural networks (DNNs) have their origins in the neural circuits of the biological brain. Recently, it has been shown that there is more to their correspondence than that. A representative example is that the brain activity produced by the presentation of a stimulus is predictable from the response of a DNN to the same stimulus using a simple linear transformation. This suggests the possibility that DNNs can be regarded as quantitative brain models implemented on a computer. While inspecting the actual brain is technically and ethically limited, DNNs can be analyzed in any way. Therefore, if DNNs can be regarded as brain models, we can deepen our understanding of the real brain through this analysis. Indeed, by using DNNs as a brain model, which is mapped to the brain by a linear transformation, new technologies such as methods of reconstructing sensory input and mental images from brain activity have been developed. Moreover, several findings suggest that efficient methods to intervene in brain activity and functions using sensory and electrophysiological stimuli may be designable by utilizing such brain models as the simulators of the brain. In the previous studies, the correspondence between the human brain and DNNs has been investigated mainly by using fMRI measurement during presenting static images as sensory stimuli to the participants. However, such conventional studies are not sufficient, considering that sensory input in natural situations is dynamic and brain information processing is also manifested in millisecond-scale activity. Therefore, in the present research, we investigated rapid dynamics induced by the presentations of the naturalistic video with auditory sounds using the electroencephalogram (EEG). We showed 10-second short clips of various scenes and 20-minute foreign-language documentaries. Then, we compared the responses of the brain to those in a DNN that was pretrained to extract representations of the short video clips. We found that the predictability of the brain activity from the DNN responses was dependent on the band frequency. Especially, the beta and gamma activity are more predictable than the theta and alpha activity. This result is physiologically plausible since the high-frequency activity is more related to the active and vigilant sensations. On the other hand, if we plan to use the DNNs as the brain model, the prediction especially of the low-frequency activity needs to be improved in future research. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-030 複数のダイポール表現を用いたMEG心拍アーチファクト除去の性能向上 Multiple dipole representation of the heart improves the performance of MEG artifact removal *森重 健一(1)、長谷川 隆成(1) 1. 富山県立大学 *Ken-ichi Morishige(1), Ryusei Hasegawa(1) 1. Toyama Pref Univ, Toyama, Japan Keyword: MEG, artifact, hierarchical Bayesian method The measurement of magnetoencephalography signals can be contaminated by large magnetic artifacts, such as heart beats, eye movements, and so on. In order to remove these artifacts, we propose an extra-dipole method for cortical current source estimation. This method is based on a hierarchical Bayesian method and simultaneously estimates the cortical and extra-brain source currents. In our previous studies we separately acquired Head/Neck/Chest T1 images in a volumetric interpolated breath-hold sequence using peripheral pulse gating and then merged to create a single large image to provide the relative location of the brain and heart. Twenty-four dipoles (eight grid points × three directions) were located around the heart, and the estimated heart currents could express a P-QRS-T sequence, and the estimated cortical currents exhibited reasonable spatial-temporal patterns. Here, we attempted to reconstruct the time series of experimental task information (attentional target positions and velocities during a covert visual pursuit task) from the estimated cortical currents using a sparse regression method and compared to the other denoising methods. To evaluate the performance of regression, four types of the coefficients were calculated. First method is “VBMEG”, that only located the dipoles on the cortical surface and estimated the dipole currents. Second one is “PCA-pruned”, that estimated dipole currents from MEG data which artifacts components were removed by using PCA. Third one is “ICA-pruned”, that estimated from MEG data which artifact components were removed by using ICA. Last one is “Extra-Dipole”, that places dipoles not only on cortical surfaces but also on extra-brain sources, and simultaneously estimated the dipole currents. We investigated which denoising method was most effective to reconstruct the target information from the estimated cortical currents. A randomized block design ANOVA found a significant statistical differences among them. Tukey’s HSD comparison showed that the correlation and determination coefficients of Extra-Dipole Method was significantly larger than the others. These results indicate that this approach was more reasonable and effective than existing methods for removing artifacts contaminating MEG signals. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-031 新しい学習環境下における神経スケールフリーネットワークの消失 Loss of Scale-free Neuronal Network in Novel Events *船橋 邦夫(1,2)、ガンドール カレド(1,2)、井ノ口 馨(1,2) 1. 富山大学医学部、2. 富山大学 アイドリング脳科学研究センター *Kunio Funahashi(1,2), Khaled Ghandour(1,2), Kaoru Inokuchi(1,2) 1. Grad Sch Med, Univ of Toyama, Toyama, Japan, 2. Research Center for Idling Brain Science, Univ of Toyama Keyword: SCALE FREE NETWORK, GRANGER CAUSALITY, Ca2+ Imaging We encounter new situations in our daily life. In novel events, we frequently feel that time is little longer than during re-exposures. This could be attributed to the numerous neuronal firings occurring that reflect the sensory inputs. However, neuronal dynamics and their characteristics’ during novelty is still poorly understood. Using a Ca2+ imaging system that combines both labelling and visualizing engram and non-engram cells in freely moving mice. We observed the neuronal activity of CA1 cells during pre-learning sleep, novel contextual exploration, post-learning sleep, retrieval, and different context sessions, via a miniature microscope. We tried three analytical methods, Granger causality network analysis, entropy calculation and theta/delta wave ratio analysis. Granger causality analysis is used to examine the directed causality (←, →, ↔) between the measured cells and create a network to evaluate it. We checked whether network type (↔) is scale free network or not and compared the network structure of each recorded session. Scale free network distribution degree follows a power law. It is widely known in the field of network science that the characteristics of scale free network are “growth” and “preferential attachment”. Scale-free network is a network structure that have hubs. Node tends to connect hubs. Interestingly, in novel situation, the network type (↔) changed from scale free network to non-scale free network. Thus, network structure changed to a no-hub structure, only during novel context exploration. Next, we calculated entropy of normal distribution and Markov model in each activity session. Calculated entropy becomes significantly higher in the first novel context than all other recorded sessions. Finally, in theta dominance analysis, we used energy ratio of theta wave and delta wave using digital filtering technique. We proposed an index, theta/delta wave ratio, and succeeded in evaluating the properties of theta dominance in 20Hz Ca2+ imaging system. The characteristics of theta dominance depends on delta wave’s fluctuation than theta waves. We propose a new analysis method in the viewpoint of network science, information theory, digital signal processing. These methods will be a useful tool for future research in assessing neural activity. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-032 異常タンパク蓄積に関連する神経変性疾患バイオマーカー評価のための階層ベイズモデル A hierarchical Bayesian model for evaluating biomarkers of neurodegenerative diseases according to the relevance to abnormal protein accumulation. *矢田 祐一郎(1)、本田 直樹(1,2,3,4) 1. 広島大学統合生命科学研究科、2. 京都大学大学院生命科学研究科、3. 自然科学研究機構 生命創成探究センター、4. 広島大学脳・こころ・感性科学研究センター *Yuichiro Yada(1), Honda Naoki(1,2,3,4) 1. Grad Sch Integrated Sciences for Life, Hiroshima University, Higashi-Hiroshima, Japan, 2. Grad Sch Biostudies, Kyoto University, Kyoto, Japan, 3. Exploratory Research Center on Life and Living Systems (ExCELLS), National Institutes of Natural Sciences, 4. Center for Brain, Mind and Kansei Sciences Research (BMK Center), Hiroshima University, Hiroshima, Japan Keyword: Alzheimer's disease, Amyloid beta, Bayesian inference The discovery of conveniently observable biomarkers of neurodegenerative diseases is an urgent issue given the growing population of the affected. Aggregation of abnormally misfolded proteins is a widely observed phenomenon in neurodegenerative diseases. Aggregated insoluble amyloid-beta plaque, for instance, accumulates in the brain of patients with Alzheimer’s disease (AD), well before the onset of deterioration of cognitive functions. The accumulation state of such proteins may represent the latent progression state of the disease. We here propose a hierarchial Bayesian regression model for evaluating biomarker candidates in model animals according to the relevance to the accumulation of abnormally misfolded proteins. In our model, the abnormal protein accumulates according to the logistic sigmoid function of time, and the observed data of biomarker candidates are generated based on the accumulation state of the protein. Due to technical difficulties, the amount of accumulated amyloid-beta is often evaluated only in the limited number of the samples that have the data of biomarker candidates. Thus, we estimate the prior distribution of the parameter of the logistic function for amyloid-beta accumulation independently and train the model in the manner of semi-supervised learning. We evaluated the model by applying it to the public data of the behavioral experiment of the AD model and wild-type mice. The predicted amount of amyloid-beta by the model correlated with the observed ground truth, which indicates the applicability of the proposed model in biomarker discovery. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-033 カルシウムイメージングデータの因子分析に対する二重ベイズ主成分分析法 Dual Bayesian PCA for Factor Analysis on Calcium imaging data *李 玉哲(1)、銅谷 賢治(1) 1. 沖縄科学技術大学院大学 *Yuzhe Li(1), Kenji Doya (1) 1. Okinawa Institute of Science and Technology Graduate University Keyword: Factor Analysis, Bayesian, automatically dimensionality reduction Factor analysis is introduced based on a latent variable model, which uses a set of latent variables to describe the observation data. It is a powerful tool for reducing the dimensions of the multivariate data, and the extracted latent variables often have meaningful interpretations. Solving a latent variable model usually requires manually choosing the latent variables' dimensionality. Bayesian PCA, a Bayesian formation of latent variable model, was proposed to select the dimensionality of the latent variables automatically. However, our applications of Bayesian PCA to calcium imaging data showed that it could not sufficiently reduce the dimensionality of the fluorescence recordings. We proposed a new method based on Bayesian PCA by adding a Bayesian formation to the latent variables, which was formulated by a zero-mean Gaussian prior, with a deviation hyperparameter as a diagonal matrix. We kept the similar settings of the factor loading matrix as Bayesian PCA, which uses the similar Baysiean formation as the latent variables; hence, we named the new model "dual Bayesian PCA." We tested our new model on simulated calcium imaging data with Poisson-Gaussian noise. Compared with other methods, the dual Bayesian PCA successfully gained an excellent sparsity in the latent variables and showed the best reconstruction performance using the dimension-reduced latent variables. We also compared the correlation of the extracted latent variables with the original data, and it showed that the reduced dimensions in latent variables kept most information from the data dimensions with higher firing rates. We also mathematically explored why the dual Bayesian PCA can gain more sparsity than Bayesian PCA on non-Gaussian data. Our proof shows that when the observation data contains much non-Gaussian noise, the dual Bayesian PCA is more accessible to the conditions for the deviation hyperparameters taking a stationary point at infinity. Therefore, it is easier for some dimensions in the latent variable to concentrate at its zero mean. Furthermore, we applied our new model to a two-photon calcium imaging data acquired from the posterior parietal cortex of mice. The results showed an excellent sparsity, and the low-dimensional latent variable captured the motion features during locomotion and the directions of the auditory stimulus. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-034 母性因子がミクログリアを介して子の脳の発達に与える影響の解析 Analysis of the effect of maternal factor on the brain development of the child via microglia *定方 瑞樹(1)、高雄 啓三(2)、金子 涼輔(3)、飯島 崇利(4)、定方 哲史(5) 1. 群馬大学医学部医学科、2. 富山大学学術研究部医学系、3. 大阪大学大学院生命機能研究科、4. 東海大学医学部医学科基礎医学系、5. 群馬大学大学院医学系研究科 *Mizuki Sadakata(1), Keizo Takao(2), Ryosuke Kaneko(3), Takatoshi Iijima(4), Tetsushi Sadakata(5) 1. Faculty of Medicine, Gunma University, 2. Life Science Research Center, University of Toyama, 3. Laboratories for Integrated Biology, Graduate School of Frontier Biosciences, Osaka University, 4. Institute of Innovative Science and Technology, Tokai University, 5. Education and Research Support Center, Gunma University Graduate School of Medicine Keyword: microglia Maternal antibodies are transferred to the blood of the child via the placenta and breast milk, and it is thought that their main function is to enhance the immunity of the newborn. On the other hand, abnormalities in the amount of antibodies in maternal plasma have been reported to be involved in autism, and a relationship between breastfeeding and IQ has also been suggested. 　We have recently found that maternal antibodies bind to microglial cells in the brains of mouse pups only during infancy. Furthermore, when we generated genetically modified mice in which maternal antibodies were not delivered to the offspring, we observed a decrease in the number of microglial cells, an increase in the overall cell density of the cerebral neocortex, and a decrease in the number of inhibitory neurons in the cerebral neocortex. In addition to further detailed anatomical analysis, we would like to conduct RNA-Seq and behavioral analysis to clarify the effects of microglial stimulated by maternal antibodies on brain development. 　As non-inflammatory roles of microglia become increasingly recognized as critical to normal neurodevelopment, it is important to consider how dysfunction in these processes might explain the seemingly disparate findings of immune dysfunction and aberrant synaptogenesis as seen in autism spectrum disorder. 　This study reveals the changes in microglia in the immature brain in response to maternal antibodies and the regulation of their functions in normal brain development. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-035 Syntaxin 1A遺伝子の神経発現を規定する抑制性制御機構の解明 The silencing mechanism determining neuronal expression of syntaxin 1A gene *中山 高宏(1)、福冨 俊之(2)、寺尾 安生(1)、赤川 公朗(1) 1. 杏林大学医学部病態生理学、2. 杏林大学医学部薬理学 *Takahiro Nakayama(1), Toshiyuki Fukutomi(2), Yasuo Terao(1), Kimio Akagawa(1) 1. Dept Med Physiol, Kyorin University, Sch Med, Tokyo, Japan, 2. Dept Pharm Toxico, Kyorin University, Sch Med, Tokyo, Japan Keyword: SNARE, gene silencing, YY1, heterochromatin The HPC-1/syntaxin 1A (Stx1a) gene, which is involved in synaptic transmission and neurodevelopmental disorders, is a TATA-less gene with several transcription start sites. It is activated by the binding of Sp1 and acetylated histone H3 to the −204 to +2 core promoter region (CPR) in neuronal cell/tissue. Furthermore, it is depressed by the association of class 1 histone deacetylases (HDACs) to Stx1a–CPR in non-neuronal cell/tissue. To further clarify the factors characterizing Stx1a gene silencing in non-neuronal cell/tissue not expressing Stx1a, we attempted to identify the promoter region forming DNA–protein complex only in non-neuronal cells. Electrophoresis mobility shift assays (EMSA) demonstrated that the −183 to −137 OL2 promoter region forms DNA–protein complex only in non-neuronal fetal rat skin keratinocyte (FRSK) cells which do not express Stx1a. Furthermore, the Yin-Yang 1 (YY1) transcription factor binds to the −183 to −137 promoter region of Stx1a in FRSK cells, as shown by competitive EMSA and supershift assay. Chromatin immunoprecipitation assay revealed that YY1 in vivo associates to Stx1a–CPR in cell/tissue not expressing Stx1a and that trichostatin A treatment in FRSK cells decreases the high-level association of YY1 to Stx1a-CPR in default. Reporter assay indicated that YY1 negatively regulates Stx1a transcription. Finally, mass spectrometry analysis showed that gene silencing factors, including HDAC1, associate onto the −183 to −137 promoter region together with YY1. The current study is the first to report that Stx1a transcription is negatively regulated in a cell/tissue-specific manner by YY1 transcription factor, which binds to the −183 to −137 promoter region together with gene silencing factors, including HDAC. 2022年7月2日　11:00～12:00　宜野湾市民体育館　ポスター会場2 3LBA-036 視覚刺激に対する防御行動の発達と経験依存的可塑性の臨界期 Development and the critical period for the experience-dependent modulation of the defensive behaviors of mice to visual threats *鳴島 円(1)、揚妻 正和(1)、鍋倉 淳一(1) 1. 生理学研究所生体恒常性発達研究部門 *Madoka Narushima(1), Masakazu Agetsuma(1), Junichi Nabekura(1) 1. Div Homeostatic Development, National Institute for Physiological Sciences Keyword: defensive behavior, development, experience dependent plasticity Organisms exhibit specific behavior patterns in response to aversive stimuli to protect themselves and survive. Although species-specific patterns of defensive behaviors are genetically encoded, individuals’ experiences gained from their habitats affect the characteristics of defensive behaviors. In particular, sensory experiences of the same modality as the triggering stimulus expect to impact post-maturity behaviors. To understand the influence of an individual’s sensory experience on innate defensive behavior, the developmental process of the behavior must be described, and the developmental plasticity of the sensory system required to receive the aversive stimuli must be considered. Rodents demonstrate defensive behaviors such as fleeing or freezing upon recognizing a looming shadow above them. Although individuals’ experiences in their habitat can modulate the defensive behavior phenotype, the effects of systematically manipulating the individual’s visual experience on vision-guided defensive behaviors have not been studied. We aimed to describe the developmental process of defensive behaviors in response to visual threats and the effect of visual experience. We found that the probability of escape response occurrence increased three weeks postnatally, then stabilized. When the visual experience was perturbed by dark rearing (DR) from postnatal day (P) 21 for a week, the developmental increase in escape probability was clearly suppressed, and the freezing probability increased. Intriguingly, exposure to the looming stimuli at P28 reversed the suppression of escape response development at P35. We also found that modulation of visual experience from P14 or P35 affected phenotypes of defensive behaviors differently and there was a critical time window for the development of escape response. These results clearly indicate that the development of defensive behaviors to looming stimuli is affected by an individual’s sensory experience.