Emergence of retinal ganglion cell response diversity from synaptic interactions in the inner retina - a combined approach of two-photon population imaging and computational modeling

为了处理信息，神经元整合了来自大量突触前伙伴的突触输入。在视网膜中，每个视网膜神经节细胞(RGC)从视觉场景中提取~20个平行表示中的一个，并将其发送到大脑。为了达到其独特的反应模式，它整合了来自特定类型的双极细胞(BC)和无长突细胞(AC)的突触输入。这发生在视网膜转换板，即内丛状层。尽管这种高度结构化的组织片在解剖学上有很好的特征，但每个BC/AC/RGC微电路中的功能连接和整合规则在很大程度上是未知的。我们将结合在三个连续水平上获得的光学记录(突触前双极细胞终末、突触后RGC树突/胞体)和计算模型来建立视网膜内部的突触连接如何产生RGC反应的特异性。作为第一步，我们将生成BC输出通道的完整功能指纹。在此之前，我们已经在小鼠视网膜中区分了8种不同的BC类型。然而，在解剖学上已鉴定出12种类型。为了确定每种解剖类型的功能特征，我们将使用双光子成像来记录单个BCS突触终末的光驱动钙变化，以响应标准化的刺激和聚集技术。利用这些数据和从RGC Somata获得的10,000多个光学记录的现有数据，我们将使用贝叶斯推理在基于我们的实验数据选择的具有灵活非线性的线性-非线性-级联模型中映射BC和RGC类型之间的功能联系。该方法将仅限于解剖学上看似合理的连接，并将允许从ACS到RGC的功能输入的推断。我们首先将重点放在三个特定的RGC电路上，然后将方法扩展到RGC的其余部分。最后，我们将确定不同类型的RGC如何沿着树突长度整合BC和AC突触输入。为此，我们将记录选定RGC类型的不同树突节段的光诱发钙信号，并将这些测量与来自SOMA的膜片钳记录和药物操作相结合。将这些数据与基于选定RGC类型的形态重建的生物物理模型相结合，将允许理清突触前输入、形态和主动树突计算对创造RGC反应特异性的贡献。这个项目将帮助我们理解不同类型的BC提供的反应构建块是如何由每种RGC类型组合在一起的，以产生观察到的视网膜输出的多样性。我们的结果将为从功能级别到实现的电路级别的一组神经计算提供独特的观点。此外，它们还可以作为更好地了解视网膜退行性疾病背后的突触基础的起点。

To process information, neurons integrate synaptic inputs from a multitude of presynaptic partners. In the retina, each retinal ganglion cell (RGC) extracts one of ~20 parallel representations from the visual scene and sends it to the brain. To arrive at its unique response pattern, it integrates synaptic inputs from specific types of bipolar (BC) and amacrine cells (AC). This happens in the retinal switchboard, the inner plexiform layer. Although this highly structured sheet of tissue is anatomically well-characterized, the functional connections and the integration rules within each BC/AC/RGC microcircuit are largely unknown. We will combine optical recordings taken at three consecutive levels (presynaptic bipolar cell terminals, postsynaptic RGC dendrites/somata) and computational models to establish how synaptic connectivity in the inner retina gives rise to RGC response specificity. As a first step, we will generate a complete functional fingerprint of BC output channels. Previously, we have differentiated 8 functionally distinct BC types in the mouse retina. However, 12 types have been identified anatomically. To identify a functional signature for each anatomical type, we will use two-photon imaging to record light-driven calcium changes in the synaptic terminals of individual BCs in response to a standardized battery of stimuli and clustering techniques. Using this data and existing data on more than 10,000 optical recordings taken from RGC somata, we will map the functional connections between BC and RGC types using Bayesian inference in linear-nonlinear-cascade models with flexible non-linearities chosen based on our experimental data. The approach will be restricted to anatomically plausible connections and will allow inferring functional inputs from ACs onto RGCs as well. We will focus initially on three specific RGC circuits before expanding the approach to the remainder of RGCs. Finally, we will determine how different RGC types integrate BC and AC synaptic inputs along the length of their dendrites. To this end, we will record light evoked calcium signals at different dendritic segments of selected RGC types and combine these measurements with patch clamp recordings from the soma and pharmacological manipulation. Combining this data with biophysical models based on morphological reconstructions of selected RGC types, will allows to disentangle the contributions of presynaptic input, morphology and active dendritic computation for creating RGC response specificity. This project will aid our understanding of how the response building blocks provided by the different BC types are combined by each RGC type to yield the observed diversity in the retinal output. Our results will offer a unique view on a set of neural computations from the functional level to the circuit level of implementation. In addition, they can serve as a starting point for a better understanding of the synaptic basis underlying retinal degenerative diseases.