一般口演
空間ナビゲーション / 学習、記憶及び可塑性など
Spatial Navigation / Learning, Memory and Plasticity
座長：林 康紀（京都大学） |
|---|
| 2022年7月2日 16:10～16:25 沖縄コンベンションセンター 会議場B2 第5会場
3O05e1-01
物理空間と意味空間におけるナビゲーションを統合する嗅内皮質の神経表現モデル
Neural representation model of the entorhinal cortex for navigation in physical and semantic spaces
*芳賀 達也(1)、大関 洋平(2)、深井 朋樹(1)
1. 沖縄科学技術大学院大学、2. 東京大学
*Tatsuya Haga(1), Yohei Oseki(2), Tomoki Fukai(1)
1. Okinawa Institute of Science and Technology, 2. The University of Tokyo
Keyword: ENTORHINAL CORTEX, GRID CELL, LANGUAGE, REPRESENTATION LEARNING
Grid cells in the entorhinal cortex has been hypothesized to be a basis of vector-based navigation in the physical space. Recently, several experiments suggest that grid-like codes in entorhinal cortex are also formed for 2-dimensional conceptual spaces such as the length of a neck and the length of legs of presented images (Constantinescu et al., 2016); competence and popularity of individuals (Park et al., 2021). Thus, similar mechanisms for representational learning and vector-based navigation (vector-based inference in conceptual spaces) are presumably working for physical and conceptual spaces. However, computational principle behind this function of the spatial system in the brain is not known. Here we present a unified neural representation model of vector-based computation in physical and semantic space, which we call successor information. Successor information is based on a previously proposed computational model of hippocampus and entorhinal cortex, successor representation (Dayan, 1993; Stachenfeld et al., 2017). Both successor information and successor representation become spatially local representations in the 2-D physical space which corresponds to place cells in hippocampus. Furthermore, dimension reduction of these place-cell-like representations yield grid-like codes of the space which can be interpreted as grid cells in entorhinal cortex, and those representation vectors can be used for goal-directed navigation in the space. However, we found that successor information can be also applied to text data by regarding words in the text as places in the physical space. Then, dimension reduction of successor information mathematically approximates a word embedding model for natural language processing in the machine learning area, hence linear combination of the representation vectors enables analogical inference. For example, a vector obtained by linear combination “King” - “Man” + ”Woman” is close to “Queen” vector, and also, “Japan” – “France” + “Paris” gives “Tokyo” vector. Our representation model suggests a hypothesis that, by extending the grid representations in 2-D physical space in the non-trivial manner, the hippocampus and the entorhinal cortex can form representations for highly complex spaces of linguistic concepts and the vector-based computational mechanism helps both spatial navigation and conceptual inferences. | | 2022年7月2日 16:25～16:40 沖縄コンベンションセンター 会議場B2 第5会場
3O05e1-02
経験依存的に生じるセル・アンサンブルの脳領域横断的な同期活動が記憶を支える
Experience-dependent inter-regional coactivations of cell ensembles support memory.
*宮脇 寛行(1)、水関 健司(1)
1. 大阪公立大学大学院医学研究科
*Hiroyuki Miyawaki(1), Kenji Mizuseki(1)
1. Grad Sch Med, Osaka Metropolitan Univ, Osaka, Japan
Keyword: Inter-regional network, Cell ensemble replay, Fear memory, Memory consolidation
Memory that is acquired during wakefulness is consolidated in the subsequent sleep periods. Memories are supposed to be represented as combinations of active cells (engram cells), referred to as cell ensembles. Accumulating evidence suggests that cell ensembles in various brain regions are reactivated during sleep periods following memory acquisition, which has important roles in memory consolidation. However, it is still controversial whether reactivated ensembles interact across brain regions. In addition, in case the ensembles inter-regionally interact, how the interactions change through the memory processes remains to be determined. To investigate these points, we used fear memories as a model system. Since basolateral amygdala (BLA), ventral hippocampus CA1 region (vCA1), and prelimbic subregion of medial prefrontal cortex (PL) are involved in fear memories, we performed simultaneous large-scale electrophysiological recordings from these regions of freely moving fear-conditioned rats. Cell ensembles were determined with independent component analyses on spike trains during fear conditioning, and then instantaneous interactions among them were estimated in pre- and post-conditioning sleep periods. We found that BLA-PL and vCA1-PL ensemble pairs were significantly coactivated in post- but not in pre-conditioning non-REM sleep epochs. These coactivations were accompanied by fast network oscillations such as amygdalar high-frequency oscillations, hippocampal sharp-wave ripples, and prefrontal cortical ripples. Furthermore, we found that BLA-PL ensemble coactivations developed rapidly at the time of memory acquisition. On the other hand, vCA1-PL ensemble coactivation was rudimentary during memory acquisition and differentiated through the post-conditioning sleep periods. Interestingly, the coactivations reappeared during memory retrieval, and the reappeared coactivations preferentially occurred during fast network oscillation events. Lastly, we found that coactivation contributing ensembles were configured before memory acquisition in BLA and PL but not in vCA1. These findings suggest that pre-configured ensembles in various brain regions quickly capture elements of experiences, and newly developed inter-regional coactivations bind the captured information, thereby supporting memory retrieval. | | 2022年7月2日 16:40～16:55 沖縄コンベンションセンター 会議場B2 第5会場
3O05e1-03
海馬CA1細胞の活動レベルの時間・文脈依存性
The Temporal and Contextual Stability of Activity Levels in Hippocampal CA1 Cells
*林 勇一郎(1)、小早川 高(1)、小早川 令子(1)
1. 関西医科大学
*Yuichiro Hayashi(1), Ko Kobayakawa(1), Reiko Kobayakawa(1)
1. Kansai Medical University
Keyword: Hippocampus, Calcium imaging, CA1, Place cells
To produce consistent sensory perception, neurons must maintain stable representations of sensory input. However, a number of recent studies have shown that neuronal representations in the hippocampus and other cortical areas gradually change over time despite no changes occurring in the stimulus, environment, or behavior. In the hippocampus, both the receptive field (place field) and activity levels of each cell vary with time (Ziv et al., 2013 Nat Neurosci 16 264, Cai et al., 2016 Nature 534 115). However, contradictory to these studies, several studies suggested that the activity levels of hippocampal cells are stable with time (Lee et al., 2020 Cell 183 620, Mizuseki and Buzsaki 2013 Cell Rep 4 1010). In the present study, we show how these two contradictory observations can be reconciled. Using genetically-engineered calcium indicator GCaMP6f and a custom head-mount fluorescent microscope, we repeatedly recorded the place cell activity of mouse hippocampal CA1 area in four different environments with different sizes and colors. The calcium trace of each cell was extracted from the fluorescence image stream using CNMF-E (Zhou et al., 2018 eLife 7 1). The firing rate of each neuron was estimated from its calcium trace using OASIS software (Pnevmatikakis et al., 2016 Neuron 89 299). The activity level of hippocampal neurons fluctuated greatly in one environment but was more stable when the activity in the four environments was averaged. Each cell’s preferred environment was frequently changed over time. The number of environments in which a cell showed activity varied across cells, ranging from zero to four, and the value also tended to be preserved. Cells that showed place cell activity in many environments had higher mean firing rates and higher spatial information content, while cells active only in a small number of environments were suitable for coding environmental identity. These results suggest that although the activity level in a single environment appears to be unstable, each cell has an inherent activity level that may play a characteristic role in the coding of space. | | 2022年7月2日 16:55～17:10 沖縄コンベンションセンター 会議場B2 第5会場
3O05e1-04
許し判断に対する皮質および皮質下脳領域の寄与-fMRI研究-
Contribution of cortical and subcortical brain areas to forgiveness judgments -fMRI study-
*平石 博敏(1)、伊東 繁(2)、尾内 康臣(1)
1. 浜松医科大学 生体機能イメージング研究室、2. 浜松光医学財団 浜松PET診断センター
*Hirotoshi Hiraishi(1), Shigeru Ito(2), Yasuomi Ouchi(1)
1. Hamamatsu University School of Medicine, 2. Hamamatsu Medical Photonics Foundation, Hamamatsu PET Imaging Center
Keyword: forgiveness, fMRI, brain activity, subcortical
Forgiveness is defined as a process of reducing one’s negative motivations toward a transgressor and restoring one’s positive motivations regarding a transgressor. Previous studies investigating neural activation concerning forgiveness have found activities of the lateral PFC, ACC, medial PFC, TPJ, insula, and precuneus which play a vital role in cognitive control. Forgiveness is important for us not only to live positively in our daily lives personally but also to keep good relationships between others within our society and forgiveness judgments and social cohesion also activate left SFG, OFC, precuneus, PCC. Because brain networks used for forgiveness and unforgiveness are different, we compared brain activities between forgivable and unforgivable judgments. 24 right-handed university students (10 females, mean21.7±1.04yrs) who gave written informed consent participated. fMRI scans were conducted with a 3-Tesla scanner. The participants were asked to judge how forgivable the behavior of a protagonist in a series of three frame picture video clips by pressing a button from -3 to +3 during fMRI scan. Depending on the judging scores as forgivable, neutral, and unforgivable, we compared brain activation results. As our results, the judging forgivable activated a region near the left para-hippocampus area (PHC). The judging unforgivable activated a region near the right Caudate area, the right Agranular retro-limbic area (BA30), and a region near the left visual association area. The judging forgivable did not activate any brain areas stronger than the judging unforgivable. The judging unforgivable activated the primary sensory (PS), the superior temporal gyrus (STG), thalamus, putamen, insula, and the DLPFC in the right hemisphere stronger than the judging forgivable (peak-level, uncorr p<0.001). Forgiveness showed that because there is a suggestion that the PHC may be involved in recognizing a paralinguistic speech profile as abnormal, leading to interpretive processing by the temporal poles and right medial frontal pole that identifies the social context as sarcastic, and recognizes the speaker’s paradoxical intentions, PHC may work for social information processing for forgiveness judgment. The unforgiveness showed that BA30 is one of the sub-regions of the retrosplenial cortex and it is known that it plays a key role in various cognitive functions. Because the BA30 had fiber connections with the hippocampus, thalamus, and prefrontal cortex includes the visual cortex, there is a suggestion that BA30 mainly received and processed scene information from the visual cortex. An unforgiveness – forgiveness comparison showed that putamen and insula are parts of the “hate circuit” and the thalamus is a gateway to mental representation. In conclusion, as previous studies reported, subcortical brain areas work for forgiveness judgment as same as cortical areas. | |
|
|
