How is Performance Evaluation Encoded in the Brain?

多巴胺（DA）-基底神经节（BG）回路对运动控制和学习至关重要。我们目前对这些回路的理解主要来自对动物学习外部奖励（如食物或果汁）的研究。然而，帕金森氏症、亨廷顿舞蹈症和肌张力障碍等疾病的症状包括与寻求奖励无关的运动行为的退化。事实上，我们的大多数行为，比如学习一项运动或一种乐器，并不是为了追求奖励而采取的简单行动，而是通过将表现与内部目标相匹配而学习到的复杂运动序列。这种类型的运动学习的机制知之甚少。鸣鸟模型提供了一个独特的机会，研究如何内部指导运动序列的建设。首先，鸟鸣是通过将一个复杂的声音序列与教师歌曲的记忆或“模板”相匹配来学习的。第二，歌曲学习需要一个DA-BG回路，这是一个易处理的“歌曲系统”的一部分。我们将利用这些优势来破译运动序列是如何在“自然”的试错过程中学习的。 为了测试DA是否在内部引导的性能评估过程中编码错误，我将记录BG投射VTA神经元，因为我使用失真的听觉反馈（Aim 1，K99相位）在特定的歌曲音节中诱导听觉错误。初步结果表明，DA编码性能误差，实际和预测性能之间的差异。DA活动在扭曲的音节后被阶段性抑制，与比预测的结果更糟一致，并且在歌曲的精确时刻被阶段性激活，当预测的扭曲没有发生时，与比预测的结果更好一致。接下来，我将解决悖论（目标2，K99阶段）的DA活动如何评估过去的行为学习，并通过重新编码BG投射VTA神经元调节正在进行的运动变异性，因为鸟类从单独唱歌（可变的“练习模式”）到唱歌给女性（刻板的“表演模式”）。最后，我将开发光学技术（目标3.1），长期监测腹侧被盖区神经元在学习相关的时间尺度，以确定DA性能错误的起源和后果（目标3.2和3.3，R 00阶段）。 我的导师Jesse Goldberg博士，共同导师Joseph Fetcho博士，以及合作者Melissa Warden博士，Chris Schaffer博士和Nozomi Nishimura博士都在钙成像和光遗传学方面拥有丰富的专业知识。发展这些技术，沿着频繁的数据展示，参加研讨会和专业课程，以及与强大的合作康奈尔神经科学社区的密切互动，将使我具备向独立过渡的必要技能。在独立的R 00阶段，我将使用这些获得的技能和创新的行为，光学和计算技术来完成所提出的目标（目标3.2和3.3），并建立一个独立的研究计划，专注于自然运动序列学习的神经机制。

Dopamine (DA)-basal ganglia (BG) circuits are critical for motor control and learning. Our current understanding of these circuits comes largely from studies of animals learning for external rewards such as food or juice. Yet symptoms of diseases such as Parkinson’s, Huntington’s and dystonia include degradation of motor behaviors unrelated to reward seeking. In fact most of our behaviors, such as learning a sport or an instrument, are not simple actions in pursuit of rewards but are instead complex motor sequences learned by matching performance to internal goals. Mechanisms of this type of motor learning are poorly understood. The songbird model offers a unique opportunity to study how internally guided motor sequences are constructed. First, birdsong is learned by matching a complex vocal sequence to the memory of a tutor song, or ‘template’. Second, song learning requires a DA-BG circuit that is part of a tractable ‘song system.’ We will leverage these advantages to decipher how motor sequences are learned during ‘natural’ trial and error. To test if DA encodes error during internally-guided performance evaluation, I will record BG-projecting VTA neurons as I induce auditory error in specific song syllables using distorted auditory feedback (Aim 1, K99 phase). Preliminary results suggest DA encodes performance error, the difference between actual and predicted performance. DA activity was phasically suppressed after distorted syllables, consistent with a worse-than- predicted outcome, and was phasically activated at the precise moment of the song when a predicted distortion did not occur, consistent with a better-than-predicted outcome. Next I will resolve the paradox (Aim 2, K99 phase) of how DA activity both evaluates past behavior for learning and also modulates ongoing motor variability by recoding BG-projecting VTA neurons as birds transition from singing alone (variable ‘practice mode’) to singing to a female (stereotyped ‘performance mode’). Finally I will develop optical techniques (Aim 3.1) to chronically monitor VTA neurons over learning-relevant timescales to determine the origins and consequences of DA performance error (Aims 3.2 and 3.3, R00 phase). My mentor, Dr. Jesse Goldberg, co-mentor Dr. Joseph Fetcho, and collaborators Drs. Melissa Warden, Chris Schaffer, and Nozomi Nishimura all have extensive expertise in calcium imaging and optogenetics. Developing these techniques, along with frequent data presentations, attendance of seminars and professional courses, and close interactions with the strong collaborative Cornell neuroscience community, will equip me with the necessary skills for transitioning to independence. In the independent R00 phase, I will use these acquired skills and innovative behavioral, optical, and computational techniques to complete the proposed aims (Aims 3.2 and 3.3) and establish an independent research program focused on the neural mechanisms of natural motor sequence learning.