Fig. 1: Experimental design and example of the strategy set-shifting task.

A Experimental design. Rats were bilaterally injected with adeno-associated virus (AAV) expressing the excitatory DREADD hM3Dq or green fluorescent protein (GFP) as control (AAV5-syn-GFP) in the medial prefrontal cortex (mPFC) and bilateral cannulas were implanted 1 mm above the nucleus accumbens (NAc). After 2 weeks of recovery, rats underwent food pretraining for 5 days, visual-cue task (VCT) training for 1–2 days and response discrimination task (RDT) for 1 day. Rats were injected either with CNO or VEH into the NAc 20 min prior to the RDT. Five to seven days later, the fMRI experiment was conducted. In the scanner, rats underwent a baseline resting-state fMRI (rs-fMRI) scan to assess functional connectivity, followed by a CNO or VEH injection into the NAc. After 20 min, a post-injection rs-fMRI scan was performed followed by cerebral blood volume (CBV) to evaluate neural activity. The entire sequence was repeated 5–7 days later with reversed VEH or CNO injection order. Lastly, the rats were sacrificed for Immunofluorescence staining. B Example of VCT. The rat was required to respond on the lever that had a visual-cue stimulus light illuminated above it. C Example of RDT. The rat was required to cease the use of a visual-cue strategy, and instead use a spatial discrimination strategy to obtain food pellets. In this example, the rat was required to always press the left lever during the RDT. D Example of different subtypes of errors. The trials were separated into consecutive blocks of 8 trials each. A perseverative error was scored when the rat pressed the right lever when the light stimulus was on the right side in early stage of the choice sequence. However, if the proportion of such errors was less than 75% in a block (6 out of 8 trials), from that time point the error was scored as a regressive error. A never-reinforced error was scored when the rat chose the right lever without cue illuminated above it.