Abstract
Long-range axonal projections provide the foundation for functional connectivity between brain regions and are critical in the modulation of behavior. Descending projections from medial prefrontal cortex (mPFC) to various target regions regulate critical behavioral functions including decision making, social behavior and mood. While specific mPFC projections have distinct behavioral roles, individual mPFC projection neurons can also innervate multiple target regions. Yet how mPFC projection neurons divide their axons across the brain is poorly understood. In this study, we mapped the axon collaterals of mPFC neurons that project to nucleus accumbens (NAc), ventral tegmental area (VTA), or contralateral mPFC (cmPFC) in mice. We used tissue clearing and light sheet fluorescence microscopy to visualize the 3-D structure of axonal arbors across the intact brain. While machine learning can automate analysis of axons in images of cleared tissue, it is challenging to train a model that generalizes to all axonal structures because the appearance of axons varies by target region. In this study, we present DeepTraCE (Deep learning-based image Tracing with Combined-model Enhancement), a new strategy for axon segmentation and quantification in images of cleared tissue. DeepTraCE is based on the deep-learning framework TRAILMAP; it achieves highly accurate axon detection by combining multiple machine learning models that are each applied to different brain regions. Using DeepTraCE, we find that cmPFC, NAc, and VTA-projecting mPFC neurons represent largely separable classes with unique axon collaterals in cortical, olfactory, and thalamic regions, respectively.
Competing Interest Statement
The authors have declared no competing interest.