Equal contribution of place cells and non-place cells to the position decoding from one-photon imaging calcium transients

The hippocampus is presumably involved in spatial navigation, the formation of long-term memories in mammals, and the construction of a cognitive map of space using a prominent feature – place cells [1] (PCs). It is unclear how PCs’ neural codes are related to memory: turnover of neural representations could be involved in learning, while codes’ stability may support long-term storage and retrieval of memories [2]. In this work, we focus on the analysis of one-photon (1P) calcium imaging data recorded from mouse CA1 during free foraging in a two-dimensional arena with the help of a miniature microscope [3]. Calcium transients, extracted with the help of CaImAn software [4], are used to compute cells’ place fields and spatial information [5] (SI) values. Assuming Poisson statistics of spike trains we identified statistically significant PCs, showing that (19.8±13.2) % of cells have significant place fields, which is higher than the fraction yielded in a similar previous study [6]. However, the median SI value of 0.27±0.05 is lower than typically observed in electrophysiological recordings. In addition, the application of Support Vector Regression [7] (SVR) was used to decode running trajectories providing errors less than 20% of the arena diameter; the correlation between the decoded trajectory and the ground truth ranged from 0.55 to 0.81. Place cells and non-place cells equally contributed to the decoder. Decoding accuracy has drastically increased upon using signal samples located non further than one sample apart, demonstrating the nature of Ca-transients’ decay. The results indicate that PCs obtained by 1P imaging in a two-dimensional arena carry much less information about the mouse’s trajectory than observed in typical electrophysiological recordings.