Neural coding and computation using noisy oscillations

Rhythmic local field potential (LFP) oscillations are a commonly observed phenomenon in the brains of diverse animal species from mollusks to humans. This activity is thought to reflect periodic and synchronized action potential firing by a population of neurons in the vicinity of a recording electrode. In awake animals, LFP oscillations are noisy, exhibiting fluctuations in amplitude and frequency, and can synchronize across multiple recording sites in distant brain regions. The power and spatial coherence of LFP oscillations have been associated with many aspects of brain function, including sensation, attention and working memory. However, the role of spatially coherent noisy oscillations in neural coding and computation remains poorly understood. Are there specific computational advantages to coordinating noise in the brain? We address this problem here through studies of oscillation-induced synchronization of neurons in acute brain slices, the awake mouse olfactory bulb and a numerically simulated network model.;Using patch clamp recordings from cortical pyramidal neurons in vitro, we show that noisy oscillatory synaptic input encodes information about a neuron's firing rate in the precise timing of its action potentials. Because of this encoding, two neurons that receive the same noisy oscillatory input fire synchronously when both are driven at the same firing rate, but desynchronize at different firing rates. We show that this rate-specific synchrony (RSS) paradigm can support many-are-equal pattern recognition computation in an in vitro network model, and that RSS is robust to non-ideal cell and stimulus conditions, as expected in vivo. Based on this work, we propose that RSS occurs in the awake brain in the presence of spatially coherent noisy oscillations.;We test the RSS hypothesis in vivo by analyzing multi-electrode recordings of LFP oscillations and mitral cell action potentials from the awake mouse olfactory bulb (OB) during passive odor exposure. We use spike-field coherence (SFC) estimation to show that single mitral cells become entrained to odor-evoked beta-band (11-29 Hz) LFP oscillations in a graded, odor-dependent manner. When two mitral cells exhibit high SFC during the same odor trial, they also show an increased rate of synchronous action potentials. We confirm that this synchrony is firing rate-specific, and that the strength of RSS depends on SFC magnitude at the time scale of 1-2 beta oscillation cycles.;Finally, we demonstrate the feasibility of synchrony-based odor recognition in the mouse OB using a numerically simulated network model in which the noisy oscillation properties and pairwise RSS statistics are matched to those observed in vivo. We also use systematic parameter variation to show that the strength and time-evolution of RSS can be optimized in the brain by "tuning" properties of the noisy stimulus and neuron population. We conclude this work by proposing experimental strategies to directly explore the role of RSS in stimulus encoding and pattern recognition computation in behaving animals.