Combining theory and experiments to infer how recurrent and top-down connectivity in the corticothalamic circuit gives rise to V1 selectivity

为了理解计算是如何在神经回路中实现的，重要的是要澄清局部、循环连接与反馈回路的作用。在紧密的合作中，我们的项目结合了理论和实验方法，并通过测量早期视觉系统的功能反应特性来进行基于模型的突触连通性推断。在之前的资助期内，我们使用体内电生理学来测量初级视觉皮层（V1）中假定的兴奋性(E)和抑制性(I)神经元群及其来自丘脑背外侧展状核（dLGN）的输入的对比敏感性和定向选择性。我们观察到，所有dLGN和大多数V1神经元（86.2%）都是对比不变性的，这也适用于各自的神经群。基于电路的对比不变性，我们结合超线性稳定建模框架，使用方向和对比响应来推断局部突触连通性和输入概况。我们发现I种群形成了一个具有最强循环连接的子网络，而来自E子网络的预测是V1响应的最弱贡献者，这与先前的连接组学研究一致。值得注意的是，我们推断V1的前馈输入表明，dLGN投射比E子网络更能靶向抑制性V1神经元。综上所述，我们的研究结果表明，人群范围内的对比度不变性是电路布线和V1反应的重要约束。在下一个资助期内，我们将在体内实验和超线性稳定电路模型的基础上，从三个方面推进：(1)我们将进一步验证推断的V1连通性，并通过光遗传学操作研究PV+抑制中间神经元对V1定向选择性、对比度敏感性和对比度不变性的抑制作用。(2)通过将皮质-丘脑反馈作为V1的额外循环输入，我们将扩展我们的理论网络模型，以包含一些生物神经回路的复杂性。我们将通过光遗传学操作丘脑L6皮质丘脑（CT）轴突终端来约束该模型，同时测量V1回路的活性以及定向和对比的选择性。(3)比较V1皮质内抑制、CT反馈和控制条件下光遗传学操作下V1响应的时间过程，我们将研究这些关键电路元件如何影响V1特征对定向和对比选择性的时间动态。我们将在我们的网络模型中测试调优和/或增益的变化是否伴随着特定网络种群之间推断连接性的变化。

To understand how computations are implemented in neural circuits it is important to clarify the role of local, recurrent connectivity versus feedback loops. In a tight collaboration, our project combines theoretical and experimental approaches and performs model-based inference of synaptic connectivity from measurements of functional response properties in the early visual system.In the previous funding period, we used in vivo electrophysiology to measure contrast sensitivity and orientation selectivity in putative excitatory (E) and inhibitory (I) populations of neurons in primary visual cortex (V1) and its inputs from the dorsolateral geniculate nucleus (dLGN) of the thalamus. We observed that all dLGN and the majority of V1 neurons (86.2 %) are contrast-invariant, and that this also holds true for the respective neural populations. Building on the contrast invariance property of the circuit, we used orientation and contrast responses in combination with a supralinear stabilized modeling framework to infer the local synaptic connectivity and input profiles. We found that the I population forms a sub-network with strongest recurrent connections, and that projections from E sub-network are the weakest contributors to V1 responses, which is in line with previous connectomics studies. Notably, our inferred feedforward inputs to V1 indicate that inhibitory V1 neurons are targeted stronger by dLGN projections than E sub-network.Taken together, our results indicate that the population-wide contrast invariance is an important constraint for circuit wiring and V1 responses.For the next funding period, we will build on our results combining in vivo experiments and the supralinear stabilized circuit model, and will advance in 3 aspects: (1) We will further test the inferred V1 connectivity and investigate via optogenetic manipulations the functional role of recurrent inhibition contributed by parvalbumin-positive (PV+) inhibitory interneurons to the establishment of orientation selectivity, contrast sensitivity and contrast invariance in V1. (2) We will expand our theoretical network model to embrace some of the complexity of biological neural circuits, by including cortico-thalamic feedback as an additional recurrent input to V1. We will constrain this model by optogenetic manipulations of L6 corticothalamic (CT) axon terminals over thalamus, while measuring the V1 circuit activity and selectivity for orientation and contrast. (3) Comparing the time course of V1 responses during optogenetic manipulations of V1 intracortical inhibition, CT feedback and control conditions, we will investigate how these key circuit elements contribute to the temporal dynamics of V1 feature selectivity for orientation and contrast. We will test in our network model whether changes in tuning and/or gain are accompanied by changes in inferred connectivity between specific network populations.