Laminar segregation of top down and bottom up processes in primary auditory cortex.

Brains are fundamentally adapted to gathering information from the external world, and making sense of it. Some information coming through the senses occurs more frequently in conjunction with particular events or outcomes, making it useful for generating predictions. Predictions can be useful to guide behaviour towards favourable outcomes, and to resolve ambiguity in the sensory-action-outcome landscape. Sensory signals of particular modalities are initially processed within specialised areas of the neocortex that receive very direct thalamic sensory input. Sensory cortex is organised hierarchically, with more complex features being processed further away, connectivity-wise, from the sensory organs. Higher-level areas receive information from lower areas, and this processing stream starting with sensory organs is referred to as 'bottom up'. However, every higher-level area also sends massive connections back to lower areas. Such connections are referred to as 'top down', and provide lower areas with real-time predictions and emphasis that higher areas are generating in response to the bottom up input. In a well-functioning system, higher-order areas supply predictions that usually aid resolving uncertainty within the bottom up stream in lower areas. If such predictions fail, however, higher areas should update predictions in the form of detecting and learning new predictive patterns in the bottom up stream. How this is implemented on the local cortical level is unknown. We know that particular cell-layers within cortical areas associate with either bottom up or top down inputs, and that there are stereotypical patterns of activity between these layers. In order to study the mechanisms of how expectation facilitates perception, it is essential to dissociate expectation from other top-down influences like attention and reward expectancy. This is possible using behavioural assays purposefully designed for this. The goal of this study was to understand what the role of particular layers is in implementing the interplay between bottom up and top down information streams. The auditory cortex is well suited for such questions, and new methods-experimentally, behaviourally and theoretically-was used to achieve several of the intended goals. I achieved several milestones as detailed in the application. A behavioural apparatus was designed from scratch, using precision laser cut acrylic sheets in conjunction with custom in-house produced electronics for reliable behavioural control and monitoring capacities. The apparatus has been thoroughly benchmarked using a basic interaural level (ILD) task paradigm. To achieve precise stimulus delivery to the two ears, I developed a novel speaker mechanism (called pokephones) which ensures short-distance stimulus delivery reliably to the ears. This mechanism greatly reduces variability across trials and rats without imposing restraints on behaviour at large (as e.g. headphones does). I established a protocol to acutely record neural activity simultaneously across all A1 cortical cell layers under urethane anaesthesia. This allowed classification of functional cell responses to sounds with various parameters such as interaural level difference, fundamental frequency and amplitude. These experiments served as the starting point for developing a protocol involving chronic silicon probe recordings in A1. Silicon probes, however, did not yield high quality signals from sufficiently large neuron ensembles over the required time period of our behavioural paradigms despite a multitude of approaches. We attribute this to tissue damage and gliosis caused by the inherent rigidity of the silicon probe shank structures interfering with the integrity of neural tissue. On account of this setback, some key goals could not be established within the funding time frame, but will be further pursued in future work. Finally, I developed and adapted quantitative tools to study neural responses to stimuli on the population level, connecting features of stimulus encoding to brain state under anaesthesia. These quantitative tools will be essential for extracting useful features in the population data from chronic preparations in follow-up studies.

Cortical sensory neuroscience typically disregards effects of active internal processes on the population level, but behavioural relevance provides the context in which cortical computations are defined. Whereas sensory cortical areas provide exquisite ex perimental control, higher order cortical studies have optimized recording and treatment of population data in a behavioural context. Cortical areas and computations are hierarchically arranged such that functional complexity increases with distance from the senses. Expectations about the sensory environment can guide behaviour toward adaptive outcomes, implemented in a 'top-down' fashion to constrain responses on lower levels, the microcircuit mechanisms for which are unknown. Cortical microcircuits are organized according to layers that contain physiologically distinct cell classes with highly specific projection patterns, locally and long-range. Top-down and bottom-up projections are defined by distinct laminar specificity leading us to propose that ac tive integration of such processes are implemented with a strong laminar bias. Only by applying an appropriate technical approach with population decoding analysis on data from animals performing a task that disentangles top-down vs. bottom-up effects can these mechanisms become tractable to scientific scrutiny. We will use cutting edge in vivo electrophysiology techniques to acquire the laminar population data necessary to rigorously address these issues. Recordings will be made in animals performing a t ask that lets us systematically manipulate prior expectations about the sensory environment. Multiple-approach population analyses will be used to determine the exact laminar contributions to top down influences. Finally, targeted pharmacogenetic manipula tions will be used to establish the causal relationship between a specific cortical layer and top down expectations. This work will shed light on a fundamental but elusive circuit-level cortical operation.