探討小腦神經空間編碼如何影響中腦多巴胺訊息及酬賞系統

Abstract

在生活中，酬賞最大化被視為決策最具體的目標，它讓我們面對環境變化時得以採取相應的行動。因此，了解腦內的神經元機制如何處理酬賞經驗是極重要且需解答的課題。在過往的研究中，中腦多巴胺神經系統被認為是酬賞訊息處理中樞，它會直接以增加神經元放電的形式編碼酬賞。而有研究表明小腦也參與在酬賞處理系統當中：小腦深部核的神經元能夠直接支配中腦腹側被蓋區的多巴胺神經元。然而，我們對於這個酬賞迴路所知仍然甚少，其神經編碼如何直接處理酬賞訊？如何影響多巴胺神經元放電？以及如何改變得到酬賞後的相應行為的因果關係仍然欠缺說明。在本研究中，我們首先發現了小腦深部核的神經元能夠用一種特定的空間編碼方式來處理酬賞訊息。我們利用小鼠酬賞型決策行為模型與活體電生理紀錄來揭密小腦如何編碼酬賞。我們發現，在酬賞出現的當下，小腦深部核的神經元群體出現了空間上同步性，而非依靠單一神經元增加放電頻率來編碼。實驗進一步發現在此小腦至中腦的迴路中，局部場電位和軸突末端的鈣離子訊號都同時反映了酬賞編碼。最後，我們應用光遺傳學及直流電注射的方式改變小腦深部核神經元的空間同步性探討神經元與行為的因果關係。有趣的是，放電同步性足夠直接的影響酬賞、並立即改變決策行為。綜合以上研究結果，我們除了支持了該迴路能夠直接處理酬賞訊息；更重要的，我們提出了一種新型小腦神經元的空間放電機制，說明了它是如何調控多巴胺酬賞系統。

Reward-based learning is a fundamental mechanism to optimize behavioral responses to the rapidly changing outside world. Mounting evidence has linked the cerebellar contribution to the midbrain dopaminergic reward system, which directly regulates rewarding behaviors. However, the precise neuronal coding of deep cerebellar nuclei (DCN)-to-midbrain ventral tegmental area (VTA) projection and how it regulates reward-driven behaviors remained unclear. Here we report that reward prediction error (RPE) was carried by DCN-to-VTA projection by measuring axonal terminal calcium dynamics as mice performed a real-time two-choice foraging task. Optogenetic inhibition of DCN-to-VTA projection dampened the RPE coding and disturbed reward driven choice behaviors. To elaborate how and why, we measured the firing of DCN neurons by in-vivo electrophysiological recording. We report that neurons in the DCN encode RPE with spatial synchrony under single trial level. We observed a precise instantaneous spatial resetting of DCN neuronal ensembles under reward RPE encoding, but not an increase of DCN single-unit firing rate. Time-frequency features of local field potentials (LFPs) in the DCN show that low frequency LFPs exhibit strong oscillation exclusively during reward encoding period, supporting the instantaneous spatial synchrony during rewarding. Using optogenetic approaches and direct DC current injection to manipulate DCN-to-VTA projection, we found that suppression of spatial synchrony of DCN neurons decreased immediate RPE coding and was sufficient to perturb real-time choice behaviors. Dendritic calcium activities of cerebellar Purkinje cells also exhibited increased spatial synchrony in single reward trial, suggesting a potential source for generating spatial synchrony in the cerebellum. Our results provide evidence that real-time reward encoding is organized into the cerebellum through a novel spatial synchrony mechanism, directly orchestrating dynamics of dopamine reward signals.