Methods for measuring neural activity during voluntary wheel running

Rodent wheel running provides a translational model to study the neurobiology of physical activity, including motivation, affect, and plasticity. However, the voluntary and unconstrained nature of wheel running makes precise behavior-to-signal alignment technically challenging.

We present a workflow that aligns fiber photometry signals with pose-derived behavior during voluntary wheel running. As a use case, we record acetylcholine activity in the ventral dentate gyrus of mature male C57BL/6 J mice and integrate pose estimation (DeepLabCut), supervised behavior classification (SimBA), spectral/event processing (FiPhA), and custom within-event trend estimations (R).

In this proof-of-concept application, acetylcholine in the ventral dentate gyrus appeared to increase 0-5 s before and throughout wheel running events during both acquisition and maintenance phases. Acetylcholine levels also showed a positive correlation with off-wheel body length.

Prior studies measuring neural activity during physical activity have relied on head-fixation or forced treadmill running, which introduce stress confounds and reduce ecological validity, or wheel rotational velocity signals, which cannot distinguish active running from passive wheel rotation without manual annotation of video frames. The present workflow addresses these limitations by using voluntary home-cage wheel running to minimize stress and supervised machine learning classification that reduces behavioral annotation time by an estimated 90% while achieving greater than 96% precision and recall.

This workflow provides a template for efficiently and accurately aligning wheel running behavior with neural in vivo signals. Our proof-of-concept demonstrates the feasibility of generalizing the approach to other neuromodulators, brain regions, and recording modalities.