Behavioural and hippocampal traces of uncertainties induced by changes in reward distributions

Charline Tessereau*, Feng Xuan*, Jack Mellor, Daniel Dombeck, Peter Dayan

In volatile environments, humans and animals face different forms of uncertainty to which they must adapt to thrive. However, our understanding of the neural basis of this adaptation is incomplete, despite, for instance, long-standing arguments about its possible dependence on neuromodulation. Here, we take advantage of the well-known spatial remapping of hippocampal place cells in the face of environmental change to interrogate these processes.

We designed a novel Uncertain Reward Task (URTask) in a virtual reality (VR) apparatus (Fig. 1b). Mice (six total, adult male) were first trained to run along a linear track in VR and lick for a water reward. Different forms of uncertainty were manipulated by varying the distributions of reward locations and changing the virtual surround (Fig. 1a). Using two-photon calcium imaging, we recorded activities of hundreds of place cells in CA1 along with mouse behaviour (licking and velocity) as uncertainty was induced and partially resolved by experience.

Given inherently variable reward locations, the place map within the reward zone fractionates, with some cells coding distance from reward, other cells firing reliably in relationship to space, and a third group being co-modulated, with reward and space-related fields (Fig. 1c). From limpid changes in reward locations, the place map undergoes a higher degree of remapping which is neither global nor rate-like (Fig. 1d). By fitting generalized linear mixed models to the calcium activity of cells and allowing the regression weights to evolve within and across runs, we will track reward- and place-modulated fields and quantify the cognitive re-mapping engendered by different forms of uncertainties.

Figure 1: a) Task: Left: Animals are trained to lick to a specific reward location every trial, then are faced with a drastic change, thus inducing ‘unexpected uncertainty’. Right: Animals are trained to lick in a constantly variable reward location, but bounded within a reward zone, thus inducing ‘expected uncertainty’. b) top: VR set-up, bottom: screenshot of one VR corridor, c) Unexpected uncertainty induces remapping. Maximum of activity after the switch (y-axis) of every cell as a function of their Maximum of activity before the switch (x-axis). d) for expected uncertainty, left: calcium activity in the standard position y-axis, right: calcium activity in the position y-axis aligned to reward. Top: cell coding distance from reward, Middle: cell firing reliably in relationship to space, Bottom: co-modulated cell, with reward and space-related fields.
Web-View
Print Page
Open in new window
Estimated DIN-A4 page-width
Go to Editor View