Data from: Interactions between circuit architecture and plasticity in a closed-loop cerebellar system

A major challenge in neuroscience is to infer the sites and directions of neural plasticity that underlie learned changes in behavior. In particular, abundant feedback pathways in the brain impede reasoning about cause and effect based on neural recording data alone. We approached this problem by studying the interactions between feedback, neural activity, and plasticity in the context of vestibulo-ocular reflex learning, a closed-loop motor learning paradigm. Our strategy was to fit a series of circuit models to a large set of neural and behavioral data. Each model differed in the strength of efference copy feedback to Purkinje cells, ranging from no feedback to very strong feedback. The primary dataset before learning was obtained from male rhesus monkeys (Macaca mulatta) trained to perform a visual fixation task, and includes neural activity from Purkinje cells in the cerebellar flocculus and horizontal eye velocity measurements in response to a wide range of vestibular and visual stimuli. Data after learning was obtained from previous publications. Whereas each model fit the extracellular recording and behavioral data, the patterns of plasticity predicted by the models fundamentally differed, with the direction of plasticity at a key site changing from depression to potentiation as feedback strength increased. We find that models with weak or no efference copy feedback to Purkinje cells are consistent with climbing fiber-driven long term depression at parallel fiber-Purkinje cell synapses and explain all experimental observations, including paradoxical changes in neural activity during a closed-loop visual task that appear to contradict the underlying plasticity. These results demonstrate how learning-related changes in neural activity can appear to contradict the sign of the underlying plasticity when either internal feedback or feedback through the environment is present.

神经科学领域的一大核心挑战，在于推断介导行为习得性变化的神经可塑性（neural plasticity）位点与方向。具体而言，大脑内存在大量反馈通路，仅依靠神经记录数据难以开展因果关系推演。我们选取前庭眼反射（vestibulo-ocular reflex）学习这一闭环运动学习范式作为研究场景，通过探究反馈、神经活动与神经可塑性之间的相互作用，攻克这一难题。本研究采用的策略为，将一系列环路模型拟合至涵盖神经与行为信息的大规模数据集。各模型向浦肯野细胞（Purkinje cells）传递的外发拷贝（efference copy）反馈强度存在差异，取值范围覆盖无反馈至极强反馈。学习前的核心数据集来自经过视觉固定任务训练的雄性恒河猴（Macaca mulatta），数据涵盖小脑绒球（cerebellar flocculus）内浦肯野细胞的神经活动，以及针对多种前庭与视觉刺激的水平眼动速度测量结果。学习后的数据集则来源于既往已发表的研究。尽管各模型均能适配细胞外记录（extracellular recording）与行为数据，但不同模型预测的可塑性模式存在本质差异：随着反馈强度提升，关键位点的可塑性方向会从长时程抑制（long term depression）转向长时程增强（long term potentiation）。本研究发现，向浦肯野细胞传递弱或无外发拷贝反馈的模型，与爬行纤维（climbing fiber）驱动的平行纤维-浦肯野细胞突触（parallel fiber-Purkinje cell synapses）长时程抑制现象相符，且能够解释全部实验观测结果，其中包括闭环视觉任务中出现的、看似与核心可塑性机制相悖的神经活动反常变化。上述结果证实，当存在内部反馈或经由环境介导的反馈时，与学习相关的神经活动变化，为何会看似与背后的可塑性方向相悖。