Sensory systems rely on statistical regularities in the experienced inputs to either group disparate stimuli, or parse them into separate categories. While considerable progress has been made in understanding invariant object recognition in the visual system, how this is implemented by olfactory neural circuits remains an open question. The current leading model states that odor identity is primarily computed in the piriform cortex, drawing from mitral cell (MC) input. Surprisingly, the role of tufted cells (TC), the other principal cell-type of the olfactory bulb (OB) in decoding odor identity, and their dependence on cortical feedback, has been overlooked. Tufted cells preferentially project to the anterior olfactory nucleus (AON) and olfactory striatum, while mitral cells strongly innervate the piriform cortex (PC). Here we show that classifiers based on the population activity of tufted cells successfully decode both odor identity and intensity across a large concentration range. In these computations, tufted cells substantially outperform mitral cells, and are largely unaffected by silencing of cortical feedback. Further, cortical feedback from AON controls preferentially the gain of tufted cell odor representations, while PC feedback specifically restructures mitral cell responses, matching biases in feedforward connectivity. Leveraging cell-type specific analyses, we identify a non-canonical feedforward pathway for odor recognition and discrimination mediated by the tufted cells, and propose that OB target areas, other than the piriform cortex, such as AON and olfactory striatum, are well-positioned to compute odor identity.
bioRxiv Subject Collection: Neuroscience