学习记忆和决策活动是动物基本的神经活动和行为。研究其神经环路基础对理解神经系统信息处理机制和相关神经系统疾病具有重要意义。蘑菇体是果蝇的嗅觉学习记忆中枢，可能与哺乳动物大脑皮层具有共同祖先。因缺少特异地高强度的γ小叶神经元Gal4品系，蘑菇体γ小叶神经元在记忆提取中的作用还不明确。本研究找到了特异性高强度标记γ小叶神经元的新Gal4品系：5-HT1B-Gal4和R92F10-Gal4。结合其它蘑菇体亚结构Gal4品系，表达Shibirets1抑制这些Gal4所标记神经元的输出，我们发现惩罚性和奖赏性早期记忆均通过γ小叶和α/β小叶神经元共同提取。果蝇蘑菇体这一记忆提取策略与哺乳动物采取用两条以上通路提取记忆的策略是保守的。Rutabaga功能缺失型突变体rut2080的挽救实验表明γ小叶神经元是蘑菇体中唯一能单独支撑依赖于Rutabaga的嗅觉早期记忆形成的神经元。阻断γ小叶神经元输出不影响嗅觉早期记忆的获取。由此推测在蘑菇体中，惩罚性和奖赏性早期记忆均主要存储在γ小叶神经元中，记忆提取时还需招募α/β小叶神经元。记忆提取后用于指导行为还需一个关键环节，即决策活动。提取的记忆信息在后续的决策过程中会受到调节和比较。我们将奖赏性和惩罚性记忆训练范式进行组合建立了AP-AV训练范式。经过一次AP-AV训练后，果蝇能同时习得和存储惩罚性记忆和奖赏性记忆。在3h时，两种记忆相互竞争性抑制，使AP-AV 3h记忆指数趋近于零。由此我们建立了基于AP-AV 3h记忆的果蝇价值决策行为模型。通过神经环路筛选发现阻断TH-Gal4标记的多巴胺能神经元不影响单独嗅觉记忆的提取，但会使多数AP-AV果蝇选择惩罚性后果。阻断章鱼胺能和5-HT能神经元不影响这一决策行为。进一步筛选发现阻断C061;MB-Gal80标记的3对PPL1多巴胺能神经元同样使果蝇决策偏向惩罚性后果，而不影响某种单独记忆的提取。通过表达dTRPA1激活C061;MB-Gal80标记的PPL1多巴胺能神经元则使果蝇决策偏向奖赏性后果。这3对PPL1神经元在蘑菇体的轴突投射集中在PPL1和PAM神经元重合的MP1区和γ2区。这暗示惩罚性和奖赏性记忆信息在这些蘑菇体亚区的输出可被调节。同时蘑菇体下游进行动作选择的决策神经元在这些亚区接收蘑菇体的记忆信息的同时，还可能接收来自PPL1多巴胺能神经元的调制。结合上述研究结果，我们提出了果蝇记忆提取和多巴胺能神经元活性双向调节记忆提取后决策行为的模型。关键词：蘑菇体；惩罚性记忆；奖赏性记忆；决策行为；多巴胺能神经元

Learning and memory, as well as decision-making, are two basic neuronal processes and behaviors in animal. The study of neural circuit supporting these two processes is crucial for understanding information processing strategy of the nervous system and the pathological basis of related neurological diseases。Mushroom body (MB) plays a central role in olfactory memory processing in Drosophila. It may share common origin with vertebrate pallium. Due to the lack of strong γ lobe-specific Gal4 line, the role of MB γ lobe neurons in olfactory memory retrieval is not clear. In addition, there is a questionable disassociation of circuits between the retrieval of olfactory early memories: aversive early memory requires output from MB α/β neurons, while appetitive early memory appears to depend on the output from MB γ neurons.We characterized two Gal4 lines strongly and specifically labeling MB γ lobe neurons: 5-HT1B-Gal4 and R92F10-Gal4. Using Shibirets1 to block the output of the neurons labeled by different MB Gal4 lines, we discovered a memory retrieval strategy similar to the two-route to retrieve memory strategy in mammals. Both aversive and appetitive early memories were retrieved through the synaptic outputs from MB γ lobe and α/β lobe neurons. By restoring the Rutabaga protein expression in rut2080 mutants, our study suggest that γ lobe is the only MB substructures capable of supporting the Rutabaga-dependent early memories. Moreover, interrupting the neuronal transmission from MB γ lobe neurons didn’t affect the acquisition of olfactory early memories. On this basis, we propose that olfactory early memory is mainly formed and stored in γ lobe neurons in MB. The retrieval of these memories requires the output from γ lobe neurons besides recruiting the output from α/β lobe neurons.After memory retrieval, decision-making is needed for instructing behavior. Decision-making requires memory retrieval. The memory information can be regulated and compared during decision-making processes. Through integrating the classic pavlovian olfactory aversive conditioning and appetitive conditioning, we built an AP-AV training paradigm. After single session of AP-AV training, flies could acquire and store aversive and appetitive memories to the same conditional stimulus simultaneously. At 3h time point, these two memories competed with each other, which made AP-AV 3h memory performance near zero. Thus, we built a valued-based decision-making model based on AP-AV 3h behavioral performance. Via a neural circuit screening, we discovered that blocking the output from TH-Gal4 labeled dopaminergic neurons did not affect the memory performance of single aversive and appetitive memories, but led most AP-AV flies prefer aversive outcome. Inhibiting the neuronal transmission from octopaminergic or serotoninerigc neurons had no impact on this decision-making behavior.A further screening suggest that blocking the output from C061;MB-Gal80 labeled 3 pairs of PPL1 DA neurons also made fly decision bias punishment outcome without affecting single memory performances. In contrast, through expressing dTRPA1 to activate C061;MB-Gal80 labeled PPL1 DA neurons made most flies prefer reward outcome. These PPL1 DA neurons mainly projected their axons to mushroom body MP1 and γ2 regions, which are the common target regions of PPL1 and PAM neurons. It indicated that the output of aversive and appetitive memories can be both regulated at these MB regions. In addition, the MB downstream neurons for action selection could be modulated by PPL1 dopaminergic neurons at these regions.Based on above results, we proposed a model integrating memory retrieval and bidirectional regulation of decision-making by DA neurons in Drosophila.Keywords：Mushroom body; Aversive memory; Appetitive memory; Decision-making; Dopaminergic neurons