シンポジウム
33 神経細胞集団による高次元情報処理と動的回路再編がもたらす記憶・認知機能の制御
33 High dimensional neural computation and network reconstruction for learning and cognitive function
座長：揚妻 正和（生理学研究所）・佐々木 拓哉（東北大学大学院薬学研究科） |
|---|
| 2022年6月30日 14:05～14:28 ラグナガーデンホテル 羽衣：中 第9会場
1S09a-01
Neural representations of learned categories in mouse prefrontal cortex
*Sandra Reinert(1), Mark Hübener(1), Tobias Bonhoeffer(1), Pieter M Goltstein(1)
1. Max-Planck-Institute for Biological Intelligence (in foundation)
Keyword: prefrontal cortex, category learning, rule switching, mouse
Grouping objects and experiences into categories is a fundamental skill for humans and many animals. Particularly, learning and recalling rules for categorization enables us to flexibly adapt to changes in context. The neuronal mechanisms that underlie the formation and switching of such rules are not fully understood. To this end, we characterized representations of learned rules in medial prefrontal cortex (mPFC) by chronically recording neuronal activity of mice learning two rules for visual categorization.
In a head-fixed go/no-go task, mice were trained to categorize drifting gratings either based on grating orientation or spatial frequency. All mice successfully learned to group the stimuli into categories and generalized the learned rule to novel stimuli. Upon a rule-switch, mice quickly remapped the stimuli onto new categories. Strikingly, the animals also generalized the new rule to stimuli they had so far only experienced with the previous rule. These behavioral findings suggest that mice, like humans and non-human primates, can learn rules for categorization.
Throughout this learning paradigm, we followed the activity of neurons in layer 2/3 of mPFC using two-photon calcium imaging. We found a set of neurons that acquired selective responses to stimuli of one of the two categories, i.e. developed category selectivity. After the rule-switch, we observed remapping of category-selective neurons, as well as previously non-selective neurons becoming responsive to the new categories. Because other task parameters, like behavioral choice and reward, can also influence prefrontal cortical activity, we aimed to disentangle the effect of those parameters on neuronal responses. By requiring mice to change their learned motor response in the task from go/no-go to go-right/go-left, we identified a set of uniquely category-modulated neurons that stably represented the category identity of stimuli, irrespective of choice and reward. Together, these results demonstrate that mouse prefrontal cortex acquires representations of learned categories and flexibly encodes changes in task rules. | | 2022年6月30日 14:28～14:51 ラグナガーデンホテル 羽衣：中 第9会場
1S09a-02
柔軟なタスクスイッチングを支える前頭前野のネットワークダイナミクス
PFC network dynamics underlying flexible task switching
*中島 美保(1)
1. 理化学研究所脳神経科学研究センター
*Miho Nakajima(1)
1. RIKEN Center for Brain Science
Keyword: PFC, Cognitive flexibility, striatum, decoder
In constantly changing environments, animals must adapt their behaviors based on informative sensory stimuli. Because the meaning of a given sensory input can vary depending on the situation, the brain must flexibly map each stimulus to its appropriate output depending on internal goals. The prefrontal cortex (PFC) is known to be key area for the implementation of such flexible transformations but how the PFC supports adaptive processing of information across multiple tasks without interference remains unclear. An important step in understanding computations in the PFC is to examine how decoders of this circuit are flexibly incorporated into their task-specific networks when the task demands shift. My previous findings demonstrated that in the rule-dependent sensory selection task, PFC neurons projecting to different territories of the tail of striatum acted to decode rule information, providing outputs that control the relative gain of meaningful and distracting inputs. These outputs were anatomically fixed, placing a fundamental constraint on the internal dynamics of the PFC. These findings led to the discovery that the requirements of specific outputs (PFC to auditory part of the striatum) to the network changed depending on the task structure. By leveraging this feature of task-dependent recruitment of specific output neurons, the current study investigates how the PFC remodels its internal network using a new task switching paradigm. This approach reveals dynamic switching of the engagement of PFC subnetworks that reflect changing task structure and suggest possible mechanisms that enable flexible behavior. | | 2022年6月30日 14:51～15:14 ラグナガーデンホテル 羽衣：中 第9会場
1S09a-03
State-space analysis for neural population dynamics
*Hideaki Shimazaki(1), Ken Ishihara(2), Ulises Rodriguez Dominguez(3), Sai Sumedh Hindupur(4), Miguel Aguilera(5), S. Amin Moosavi(6), Magalie Tatischeff(7), Jimmy Gaudreault(8), Christian Donner(9)
1. Center for Human Nature, Artificial Intelligence, and Neuroscience (CHAIN), Hokkaido University, Sapporo, Japan, 2. Graduate School of Life Science, Hokkaido University, Sapporo, Japan, 3. Mathematics Research Center, Guanajuato, Mexico, 4. Department of Electrical Engineering, Indian Institute of Technology Bombay, Mumbai, India, 5. School of Engineering and Informatics, University of Sussex, Brighton, UK, 6. Department of Neuroscience, Brown University, Providence, Rhode Island, USA, 7. Department of Physics and Astronomy, University of British Columbia, Vancouver, Canada, 8. Department of Chemical Engineering, Polytechnique Montreal, Montreal, Canada, 9. Swiss Data Science Center, ETH Zurich, Zurich, Switzerland
Keyword: Population activity, Ising model, State-space model
Co-variability of neural population activity reflects internal cognitive states of an animal, constrains information coding of sensory stimuli and behavior, and can provide insights into the mechanisms of underlying circuits. However, it is still challenging to elucidate the co-variability from spiking activities of neurons recorded from awake, behaving animals because they exhibit transient dynamics. Both individual firing rates and neurons’ co-activities can vary within experimental trials. To capture the non-stationary population dynamics, we extended a standard recurrent network model known as the Ising model/Boltzmann machine using a state-space framework. Here we introduce the current status and prospects of the research using the state-space Ising model.
With the presented method, researchers can visualize the dynamics of the firing rates of individual neurons and their interactions from spiking data. It can also reveal macroscopic features of the neural system, including sparsity and interaction strength, and test whether the system is near a critical state, using thermodynamic quantities such as free energy, entropy, and heat capacity. All of these quantities are obtained in a time-resolved manner, making it possible to interpret them in comparison with the stimulus and behavioral paradigm of the experiments. We introduce our recent effort to include asymmetric kinetic dynamics in the state-space framework to account for the non-stationary and non-equilibrium neuronal dynamics. Here, measures of the non-equilibrium processes, such as conditional entropy and entropy production, can underpin the richness of population activity and the signature of the criticality appearing near non-equilibrium phase transitions.
These data-driven methods offer valuable tools to characterize cortical states and their dynamical changes. They also elucidate population coding and underlying mechanisms when combined with appropriate encoding models. We introduce our Python-based analysis environment that automatically inspects basic spiking statistics, performs model selection, and visualizes time-varying neuronal interactions and macroscopic dynamics once users input their data. | | 2022年6月30日 15:14～15:37 ラグナガーデンホテル 羽衣：中 第9会場
1S09a-04
前頭前野における恐怖連合記憶の実装と動的神経回路再編
Functional rewiring of prefrontal cortical networks by associative learning
*揚妻 正和(1)
1. 生理学研究所
*Masakazu Agetsuma(1)
1. National Institute for Physiological Sciences
Keyword: Associative fear learning, prefrontal cortex, in vivo two photon imaging, machine learning
Animals learn to adapt to changing environments for survival. Associative learning, such as classical conditioning, is one of the simplest types of learning that has been intensively studied over the past century. During the last two decades, technical development in molecular, genetic, and optogenetic methods has enabled the identification of a population of neurons in the brain, as named the memory engram, which encodes and regulates associative memory. However, how information is stored and processed by the neural population to encode and retrieve the associative memory remains unclear. The dorsal part of the medial prefrontal cortex (dmPFC) of rodents is a brain region demonstrated to be important for the retrieval of associative fear memory. During fear memory retrieval, activated individual neurons or an enhanced synchrony of neural populations in the dmPFC are observed, while pharmacological or optogenetic silencing of the dmPFC and its projections to specific downstream targets suppresses fear memory retrieval, revealing that associative fear memory is normally stored in the dmPFC. Therefore, the dmPFC can serve as an interesting target to address the fundamental question of what structural and computational alterations in the prefrontal networks are required to organize novel associative memories. To address this point, we developed a pipeline for computational dissection and longitudinal two-photon imaging of neural population activities in the dmPFC during fear-conditioning procedures, enabling us to detect learning-dependent changes in the topology and information coding of the dmPFC. Through regularized regression methods and graphical modeling, we revealed that fear conditioning organized neuronal ensembles encoding conditioned responses (CR), with enhancing their coactivity, functional connectivity, and association with conditioned stimuli (CS). This suggests that fear conditioning drives dmPFC reorganization to generate novel neural circuits for CS-to-CR signal transformation. Importantly, we found that neurons strongly responding to unconditioned stimuli (US) during conditioning anterogradely became a hub of the CR ensemble after the memory consolidation. Collectively, we demonstrate learning-dependent dynamic modulation of population coding structured on an activity-dependent hub-network formation within the dmPFC. | | 2022年6月30日 15:37～16:00 ラグナガーデンホテル 羽衣：中 第9会場
1S09a-05
新奇学習と記憶固定における海馬の情報再生の意義の解析
The roles of awake hippocampal replays for online learning and offline memory consolidation
*佐々木 拓哉(1)
1. 東北大学大学院薬学研究科
*Takuya Sasaki(1)
1. Grad Sch Pharm, Tohoku University
Keyword: hippocampus, replay, memory consolidation, learning
As a ubiquitous learning rule to reorganize functional connections across neurons, a number of studies have supported the Hebbian postulate of “fire together, wire together”. During voluntary movement, hippocampal place cells fire at specific locations of an environments. Place cells with overlapping place fields offer appropriate opportunities to generate synchronized spikes to trigger the Hebbian plasticity. In a post-experience offline state such as rest and sleep periods, a large number of hippocampal neurons that encoded learned experiences are synchronously reactivated for memory consolidation. However, post-experience reactivation contents are not a mere replication of neuronal encoding patterns of waking experiences. The difference between awake encoding and post-experience reactivation of neuronal ensembles could be reconciled by awake replay events, coinciding with awake SWRs, that preferentially emerge during non-exploratory consummatory behavior such as reward consumption. These online replay events also provide temporal windows suited to promote the Hebbian plasticity. Here, we examined how awake hippocampal replays in a novel spatial task contribute to shaping post-task reactivation patterns. We showed that neuronal ensembles recruited by awake replays during consummatory periods, rather than the entire set of place cell ensembles, explain better than those reactivated in a post rest period. Neuronal pairs with correlated spikes during consummatory periods, in addition to spatial firing during running periods, exhibited stronger potentiation in post rest reactivation. Closed-loop inhibition of sharp wave ripples during consummatory periods substantially diminished potentiation in post rest reactivation of neuronal ensembles. These results highlight the essential role of learning-induced awake replays in the extraction of hippocampal neuronal ensembles to be potentiated for subsequent memory consolidation. | | | |
|
|
