探討思覺失調症候選基因 Akt1 及背內側紋狀體在小鼠酬賞決策過程中所扮演的角色

Investigating The Role of Akt1, A Schizophrenia Candidate Gene, and The Dorsomedial Striatum in Reward-Based Decision Making in Mice

張家源(Chia-Yuan Chang)

指導教授 ： 賴文崧

國立臺灣大學/理學院/心理學研究所/博士(2020年)

https://doi.org/10.6342/NTU202002561

摘要

思覺失調症是一種伴隨各種認知缺損的重大精神疾病，越來越多證據顯示有許多基因是思覺失調症的致病候選基因，其中包含 AKT1 在內。AKT1 本身是蛋 白質激酶也是多巴胺二型受體（D2 receptor）的下游，對於調節多巴胺訊號傳遞扮演重要角色。過去研究顯示，多巴胺訊號傳遞的變化可能參與在思覺失調症的致病機制，且病人在許多決策相關作業上表現比一般人差。近期研究更進一步指出 AKT1 在依賴紋狀體（striatum）功能的酬賞相關決策作業中是不可或缺的。 然而到目前為止，AKT1 到底在酬賞相關決策作業歷程的神經機制中，特別是在 背內側紋狀體（dorsomedial striatum)的神經機制，扮演什麼樣的角色仍不清楚。 在本研究當中，我們提出了四個研究目標去檢驗 AKT1 在酬賞決策行為中有何影響，特別針對背內側紋狀體神經活動及特定型態的神經元與決策行為之間的關係作探討。（1）AKT1 對二擇一動態酬賞作業表現的影響。（2）AKT1 在決策 行為不同階段時如何調節背內側紋狀體神經活動。（3）背內側紋狀體神經活動 與決策行為的因果關係。（4）背內側紋狀體中特定型態的神經元對決策行為的 影響。實驗結果顯示（1）相較於控制組，Akt1 異型合子小鼠只需要較少的試驗 數便可學會作業，並且勝利保持行為（win-stay）的比例也顯著地高。更利用強 化學習（reinforcement learning）的數學模型去推估小鼠決策行為的模型參數，其 顯示 Akt1 異型合子小鼠擁有較高對無酬賞結果的學習速率，但同時表現較低的 選擇一致性。（2）活體電生理記錄顯示在小鼠背內側紋狀體中，無酬賞結果所引發的事件相關神經活動與決策作業中的行為表現及決策行為的模型參數有高度相關。在利用機器學習去解碼背內側紋狀體的局部場電位（local field potential) 時，我們發現此腦區神經活動包含許多決策相關資訊在內。（3）利用化學遺傳 學方法（chemogenetic）直接抑制 Akt1 異型合子小鼠的背內側紋狀體神經活性會導致異型合子小鼠的決策行為表現與野生型小鼠相似，包含需要較多的試驗數才學會作業，且勝利保持行為（win-stay）的比例也顯著地低。（4）我們進一步利 用 GAD-cre 小鼠及 PV-cre 小鼠，專一性地操弄背內側紋狀體中 γ-胺基丁酸 （GABA）神經元及表現小白蛋白（parvalbumin, PV）的中介神經元，行為資料顯示專一性地剔除小鼠背內側紋狀體中的小白蛋白中介神經元會導致較高的勝利保持行為比例及較高的學習速率及較低的選擇一致性。 . 根據上述的結果，Akt1 基因缺損極有可能是藉由改變背內側紋狀體中的小白蛋白中介神經元的活動影響到個體策略抉擇進而改變其在酬賞相關決策作業中的表現。

關鍵字

Akt1 ； 思覺失調症 ； 強化學習 ； 二擇一動態酬賞作業 ； 背內側紋狀體 ； 小白蛋白 ； γ-胺基丁酸 ； 機器學習

並列摘要

Schizophrenia is a severe neuropsychiatric disorder in which cognitive impairment features prominently. Accumulating evidence from human genetic studies suggests that multiple susceptibility genes might contribute to the pathogenesis of schizophrenia, including AKT1 (protein kinase B α), a key signaling kinase intermediate downstream of dopamine D2 receptor. Alterations of dopaminergic transmission have been implicated in the pathogenesis of schizophrenia, and patients with schizophrenia also show worse performance than healthy controls in many decision-making tasks. Recent findings further revealed that Akt1 might play a crucial role in the modulation of reward-based decision making, especially in the striatum. However, the importance of Akt1 in reward-based decision making and its specific role in the dorsomedial striatum (DMS) during this process remain elusive. To this end, we proposed 4 aims to examine the role of Akt1 in the reward-related decision making behavior and the relationship between neural activity or specific type of neurons in DMS and behavioral performance in decision making task. (1) the impact of Akt1 in a probabilistic two-choice foraging task. (2) The role of Akt1 in the regulation of dorsomedial striatum (DMS) neural activity during different stages of decision-making. (3) The causal relationship between DMS neural activity and decision-making behavior. (4) The cell type-specific role of DMS neurons in decision-making. Our results indicated that (1) Akt1 mutant mice required significantly fewer trials to achieve the criteria and higher ratio of win-stay behavior compared to their controls in both acquisition and reversal phases. Taking advantage of a Bayesian approach to estimate the parameters in a modified reinforcement learning model, we found that Akt1 mutant mice have a higher learning rate in the no reward outcomes and lower choice consistency compared to controls. (2) The in vivo recording showed that no-reward-evoked power in the DMS is highly correlated with their behavioral performance and reinforcement learning model parameters. The decoding of local field potential in DMS by machine learning reveals the choice related information is embedded in DMS neural activity and it also shows the genotypic difference. (3) The direct inhibition of DMS by the chemogentic approach in Akt1 mutant mice make WT-like behavior, including higher accumulated trials and lower ratio of win-stay behavior. (4) The behavioral results from cell type-specific modulation in DMS through GAD-cre mice and parvalbumin (PV)-cre mice shows that the lesion of PV interneurons made the higher ratio of win-stay behavior, higher learning rate and lower choice consistency. Collectively, our results suggest that Akt1 deficiency cause the alternation of neural oscillation in the DMS potentially through the PV interneurons in DMS, which is contributed to the learning rate which resulted in the differential selection of choice strategy (e.g. win-stay) during decision making.

並列關鍵字

Akt1 ； schizophrenia ； reinforcement learning ； two choice dynamic foraging task ； dorsomedial striatum ； parvalbumin ； GABA ； machine learning

參考文獻

Abi-Dargham, A. (2004). Do we still believe in the dopamine hypothesis? New data bring new evidence. International Journal of Neuropsychopharmacology, 7 Suppl 1, S1-5. doi:10.1017/S1461145704004110

Ahn, W. Y., Haines, N., Zhang, L. (2017). Revealing neurocomputational mechanisms of reinforcement learning and decision-making with the hBayesDM package. Computational Psychiatry, 1, 24-57. doi:10.1162/CPSY_a_00002

Alptekin, K., Akvardar, Y., Kivircik Akdede, B. B., Dumlu, K., Isik, D., Pirincci, F., . . . Kitis, A. (2005). Is quality of life associated with cognitive impairment in schizophrenia? Progress in Neuro-Psychopharmacology Biological Psychiatry, 29, 239-244. doi:10.1016/j.pnpbp.2004.11.006

Aparicio-Legarza, M. I., Davis, B., Hutson, P. H., Reynolds, G. P. (1998). Increased density of glutamate/N-methyl-D-aspartate receptors in putamen from schizophrenic patients. Neuroscience Letters, 241, 143-146. doi:10.1016/s0304-3940(98)00017-2

Balan, S., Yamada, K., Iwayama, Y., Hashimoto, T., Toyota, T., Shimamoto, C., . . . Yoshikawa, T. (2017). Comprehensive association analysis of 27 genes from the GABAergic system in Japanese individuals affected with schizophrenia. Schizophrenia Research, 185, 33-40. doi:10.1016/j.schres.2017.01.003