A cell-type specific explanation of visual decision circuits.

点击翻译按钮获取中文摘要

Abstract The main goal of my laboratory is to understand the neural circuits that support perceptual decision-making. Recent efforts to understand microcircuits within decision-making areas have been fruitful, thanks in part to the ability to identify and manipulate distinct types of inhibitory neurons. Our understanding how distinct types of excitatory pyramidal neurons (PyNs) shape circuits has lagged behind. This lack of understanding is particularly problematic for decision circuits in which PyN type can determine the long-range project target of neurons within an area. To surmount this problem, we propose to compare the contribution of distinct PyN types to key decision- making computations: working memory and evidence accumulation. We will measure and manipulate neural activity of intratelencephalic (IT) and pyramidal tract (PT) neurons in animals trained to make decisions about the stochastically-varying spatial position of visual gratings. We seek to understand how PT and IT neurons in multiple neural structures collectively support the computations needed for decision-making. In Aim 1, we will use widefield calcium imaging to generate cortex-wide activity maps for each PyN type. We will compare the spatial and temporal patterns at multiple scales, and will deploy quantitative analyses to identify candidate regions for decision-making computations. In Aim 2, we will measure the activity of single cells within these regions using 2-photon imaging. We will use a generalized linear model (GLM) to estimate the extent to which single neurons are modulated by relevant variables, such as stimulus events, choice, and movements. We will compare these across PT and IT neurons. The same models will estimate the magnitude and time course of “coupling” between neurons, which we will compare across PyN types and regions. Shorter coupling time courses are expected for neurons that support sensory encoding, while longer coupling time courses are expected for neurons that support working memory. We will use these observations to generate a new, biologically realistic model that includes PyN types in multiple areas working together to support decision-making computations. In Aim 3, we will evaluate a causal role for PyN types using cell-type specific optogenetic inactivation. We will compare multiple aspects of decision-making behavior on trials with vs. without inactivation. For example, we will determine how inactivation affects decision accuracy, and the extent to which this depends on when inactivation occurs. We will also determine how inactivation impacts the animal’s ability to accumulate sensory evidence and/or drives movements that can be detected with a classifier trained on video data. We will use this data to “stress test” the model, evaluating whether inactivations in our artificial network generate the same changes as inactivations in the real brain. If not, we will update and re-test the model, creating a tight loop between experiments and modeling. Taken together, the proposed work will advance our understanding of decision-making beyond the non-specific approach that has limited a mechanistic understanding thus far.