Mark H. Histed, PhD - US grants

To understand how neuronal activity patterns give rise to behavior, we have adopted a new technology that extends a two-photon microscope to stimulate both single neurons and patterns of individual neurons. This approach opens the door to studying how animals' choices depend on patterns of neuronal activity. We have achieved stable psychophysical performance in mice in the laboratory, and collected data in which we use the instruments to evoke stimulation patterns. We hypothesize that the brain regards as similar, for behavioral decisions, many different patterns that share similar statistical structure. However, it may be possible that particular neurons or groups of neurons are exclusively observed by the brain during the decision process. Our research is directed at determining which of these hypotheses is true. The experiments will shed light on how our brains interpret the thousands or millions of neurons whose activity is constantly changing in response to our perceptions and actions. Determining how the brain uses these activity patterns, which seem to be central to much of brain function, will be an important step forward in understanding how our brains work, both in health and disease.

In one project, we reported results that shed light on input-output transformations in cortical networks. We recorded spiking responses from visual cortex (V1) of awake mice of either sex while pairing sensory stimuli with optogenetic perturbation of excitatory and parvalbumin-positive inhibitory neurons. We found V1 neurons average responses were primarily additive (linear). We used a recurrent cortical network model to determine if these data, as well as past observations of nonlinearity, could be described by a common circuit architecture. Simulations showed cortical input-output transformations can be changed from linear to sublinear with moderate (20%) strengthening of connections between inhibitory neurons, but this change away from linear scaling depends on the presence of feedforward inhibition. Simulating a variety of recurrent connection strengths showed that, compared to when input arrives only to excitatory neurons, networks produce a wider range of output spiking responses in the presence of feedforward inhibition. In a second project, we are examining how the cortex controls the gain of responses by using a learning paradigm. Animals and humans improve their performance in sensory tasks with practice. But it is not known in general whether cortical representations become stronger (response gain increases) with practice, or whether downstream decoding becomes more effective. To examine the neural basis of learning, we are training animals to perform controlled learning tasks and examining how neural responses change with learning. We will adapt the models from project 1 to predict what circuit changes might lead to any response changes, and follow up that work with experimental tests of the predictions. This work will explain how the brain, particularly the cerebral cortex, changes during learning to support improved behavior. Understanding the circuit elements that change with learning is made possible by insights from theoretical work on networks that relates neural activity to circuit and anatomical features.

Project Summary To survive, organisms must both accurately represent stimuli in the outside world, and use that representation to generate beneficial behavioral actions. Historically, these two processes ? the mapping from stimuli to neural responses, and the mapping from neural activity to behavior ? have largely been treated separately. Of the two, the former has received the most attention. Often referred to as the ?neural coding problem,? its goal is to determine which features of neural activity carry information about external stimuli. This approach has led to many empirical and theoretical proposals about the spatial and temporal features of neural population activity, or ?neural codes,? that represent sensory information. However, there is still no consensus about the neural code for most sensory stimuli in most areas of the nervous system. The lack of consensus arises in part because, while it is established that certain features of neural population responses carry information about specific stimuli, it is unclear whether the brain uses (?reads?) the information in these features to form sensory perceptions. We have developed a theoretical framework, based on the intersection of coding and readout, to approach this problem. Experimentally informing this framework requires manipulating patterns of neuronal activity based on, and at the same spatiotemporal scale as, their natural firing patterns during sensory perception. This work must be done in behaving animals because it is essential to know which neural codes guide behavioral decisions. In the first phase of this project (funded by the BRAIN Initiative), we developed the technology necessary for realizing this goal. In the present proposal, we will extend our patterned neuronal stimulation technology and apply it to answer long-standing questions about neural coding and readout in the visual, olfactory, and auditory systems. We will pioneer the capacity to determine which neurons within a network are encoding behaviorally relevant information, and also to determine the extent to which temporal patterns of those neurons? activity are being used to guide behavior. Finally, we will study these neural coding principles across changes in behavioral state and during learning to determine how internal context and past experience shape coding and readout. The contributions of the proposed work will be three-fold. First, we will provide the neuroscience community with the tools needed to test theories of how neural populations encode and decode information throughout the brain. Second, we will reveal fundamental principles of spatiotemporal neural coding and readout in the visual, olfactory, and auditory systems of behaving animals. And third, our unifying theoretical framework for cracking neural codes will allow the broader neuroscience community to resolve ongoing debates regarding neural coding that have been previously stalemated by considering only half of the coding/readout problem.