UC Berkeley

The brain contains neurons of many different types interacting in complex functional circuits. To process sensory information these cells work in concert to form representations of the external world. In the auditory cortex, this involves integrating information from different cell types across an orderly anatomical structure of layers and columns. Representations can be observed at the level of single cells, cortical microcircuits, and large-scale sensory maps. The relationship between single cell properties and circuits within the auditory cortex, however, is still poorly understood. Furthermore, the structure-function relationships uncovered by neuroscientific study may crucially depend on the stimuli used to probe the system. This thesis brings together work from each of these different levels to describe how sounds are represented in the cortex, how this representation changes with experience, and how different cells contribute to cortical representation.

First, I describe how the statistics of sound stimuli influence response properties in the mouse primary auditory cortex by comparing responses to pure tones and natural sounds (ultrasonic vocalizations). I also compare these responses to a temporally reversed vocalization to determine whether a sound with similar spectrotemporal content but no ethological relevance is represented similarly. When comparing pure tones and vocalizations, I find that the temporal response properties are similar, but that spectral response properties (e.g. frequency selectivity) often differ substantially. In particular, there are multiple sites that responded to vocalizations with frequency content outside their classical tone-derived receptive field, suggesting some specificity for behaviorally relevant sounds. When comparing forward and backward vocalizations, temporal responses are similar, but frequency bandwidth and characteristic frequency differs significantly across the population. Thus, the behaviorally relevant sound appears to be represented differently from non-behaviorally relevant synthetic and naturalistic sounds.

The response properties of auditory neurons are not fixed, but rather depend on experience. In the next study, I examine how exposure to pulsed noise during different sensitive windows of the auditory critical period affects single site properties as well as circuit-level dynamics. On the single site level, I find that early exposure to pulsed noise increases receptive field thresholds and decreases frequency selectivity, while late noise exposure increases frequency bandwidths as well as spontaneous and evoked firing rates. To describe changes in functional microcircuits, I use the Ising model, which describes pairwise interactions between simultaneously recorded sites in the auditory cortex as well as interactions between sites and the stimuli that modulate them. I find that early noise exposure decreases stimulus drive, whereas late noise exposure does not change the strength of sound inputs but rather decreases the spread of functional connections from the deep to the superficial layers across sites with different frequency selectivity.

Finally, I use a combination of optogenetic tools and computational methods to describe how the activity of a specific class of inhibitory neurons affects network connectivity in the auditory cortex. I examine the contribution of parvalbumin-positive (PV+) inhibitory interneurons, which make up around half of the inhibitory neurons in the cortex. These neurons are known to be involved in the generation of gamma oscillations, and their maturation corresponds with the end of the auditory critical period for plasticity. Using Ising models in tandem with linear-nonlinear vector autoregressive models, I show that stimulating PV+ neurons increases feedforward information flow through cortical circuits without changing lateral interactions within the same layers.