Dopamine Modulates Adaptive Prediction Error Coding in the Human Midbrain and Striatum.
Dopamine Modulates Adaptive Prediction Error Coding in the Human Midbrain and Striatum.
复制标题
DOI：
10.1523/jneurosci.1979-16.2016
复制
发表时间：
2017-02-15
期刊：
影响因子：
--
通讯作者：
Fletcher PC
中科院分区：
文献类型：
--
作者：
Diederen KM;Ziauddeen H;Vestergaard MD;Spencer T;Schultz W;Fletcher PC
学习最佳地预测奖励需要代理考虑奖励值的波动。最近的研究表明，个人可以通过调整学习率以及编码与奖励可变性相关的预测误差来有效地了解可变奖励。这种适应性编码与非人类灵长类动物的中脑多巴胺神经元有关，并且从功能磁共振成像数据中浮现出支持人类多巴胺能系统具有类似作用的证据。在这里，我们试图利用多巴胺能激动剂（溴隐亭）和拮抗剂（舒必利）进行受试者间安慰剂对照药理学功能磁共振成像研究，研究多巴胺能扰动对人类自适应预测错误编码的影响。参与者执行了一项先前验证的任务，在该任务中，他们预测了从具有不同 SD 的分布中获得的即将到来的奖励的大小。每次预测后，参与者都会获得奖励，从而产生逐次尝试的预测错误。在安慰剂下，我们复制了之前对中脑和腹侧纹状体适应性编码的观察结果。舒必利治疗减弱了中脑和腹侧纹状体的适应性编码，并与表现下降相关，而溴隐亭则没有显着影响。尽管我们观察到 SD 对各组之间的表现没有差异影响，但计算模型表明舒必利组的行为适应能力下降。这些结果表明，正常的多巴胺能功能对于自适应预测错误编码至关重要，这是大脑的一个关键特性，被认为有助于在可变环境中实现高效学习。至关重要的是，这些结果还为了解多巴胺功能紊乱对精神疾病的影响提供了潜在的见解。意义陈述 为了做出最佳选择，我们必须了解会发生什么。当奖励结果存在很大的可变性时，人类的学习就会受到抑制，而受大脑化学物质多巴胺调节的两个大脑区域对奖励的可变性很敏感。在这里，我们的目标是将多巴胺与学习可变奖励以及相关教学信号的神经编码直接联系起来。我们使用多巴胺能药物扰乱健康个体的多巴胺，并要求他们在进行脑部扫描时预测可变的奖励。多巴胺扰动损害了学习和奖励变异性的神经编码，从而在多巴胺和奖励变异性适应之间建立了直接联系。这些结果有助于我们了解与多巴胺能功能障碍相关的临床状况，例如精神病。
Learning to optimally predict rewards requires agents to account for fluctuations in reward value. Recent work suggests that individuals can efficiently learn about variable rewards through adaptation of the learning rate, and coding of prediction errors relative to reward variability. Such adaptive coding has been linked to midbrain dopamine neurons in nonhuman primates, and evidence in support for a similar role of the dopaminergic system in humans is emerging from fMRI data. Here, we sought to investigate the effect of dopaminergic perturbations on adaptive prediction error coding in humans, using a between-subject, placebo-controlled pharmacological fMRI study with a dopaminergic agonist (bromocriptine) and antagonist (sulpiride). Participants performed a previously validated task in which they predicted the magnitude of upcoming rewards drawn from distributions with varying SDs. After each prediction, participants received a reward, yielding trial-by-trial prediction errors. Under placebo, we replicated previous observations of adaptive coding in the midbrain and ventral striatum. Treatment with sulpiride attenuated adaptive coding in both midbrain and ventral striatum, and was associated with a decrease in performance, whereas bromocriptine did not have a significant impact. Although we observed no differential effect of SD on performance between the groups, computational modeling suggested decreased behavioral adaptation in the sulpiride group. These results suggest that normal dopaminergic function is critical for adaptive prediction error coding, a key property of the brain thought to facilitate efficient learning in variable environments. Crucially, these results also offer potential insights for understanding the impact of disrupted dopamine function in mental illness. SIGNIFICANCE STATEMENT To choose optimally, we have to learn what to expect. Humans dampen learning when there is a great deal of variability in reward outcome, and two brain regions that are modulated by the brain chemical dopamine are sensitive to reward variability. Here, we aimed to directly relate dopamine to learning about variable rewards, and the neural encoding of associated teaching signals. We perturbed dopamine in healthy individuals using dopaminergic medication and asked them to predict variable rewards while we made brain scans. Dopamine perturbations impaired learning and the neural encoding of reward variability, thus establishing a direct link between dopamine and adaptation to reward variability. These results aid our understanding of clinical conditions associated with dopaminergic dysfunction, such as psychosis.