Behavioral experience and flexibility are crucial for survival in a constantly changing environment. Despite evolutionary pressures to develop adaptive behavioral strategies in a dynamically changing sensory landscape, the underlying neural correlates have not been well explored. Here, we use genetically encoded voltage imaging to measure signals in primary somatosensory cortex (S1) during sensory learning and behavioral adaptation in the mouse. In response to changing stimulus statistics, mice adopt a strategy that modifies their detection behavior in a context dependent manner as to maintain reward expectation. Surprisingly, neuronal activity in S1 shifts from simply representing stimulus properties to transducing signals necessary for adaptive behavior in an experience dependent manner. Our results suggest that neuronal signals in S1 are part of an adaptive framework that facilitates flexible behavior as individuals gain experience, which could be part of a general scheme that dynamically distributes the neural correlates of behavior during learning.

Behavior set up was programmed as a custom software written in Matlab and Simulink (Ver. 2015b; The MathWorks, Natick, Massachusetts, USA), which is available upon request. Voltage fluorescent data was collected using MiCAM05-N256 (Scimedia, Ltd) software. 

Open-source software was used for Image data processing (MiCAM05-N256, Scimedia, Ltd). Psychometric data analysis and curve fits were performed using open-source software psignifit toolbox version 2.5.6 for MATLAB version 5 and up (Wichmann & Hill, 2001 a,b). All other analyses were performed with custom MATLAB programs, which is available in “Dryad” with the identifier https://doi.org/10.5061/dryad.h18931zmm.

In order to access the data and computer code that support the findings of this study, all files can be downloaded from Dryad. Files are compressed ZIP files (Windows 10) and need to be saved and extracted before access.

The main dataset contains GEVI (genetically encoded voltage imaging) data from primary somatosensory cortex (S1) and simultaneously recorded behavioral readouts (licks and whisker stimulation) in a Go-No-go detection task.

Open-source software was used for image data processing (MiCAM05-N256, Scimedia, Ltd). Psychometric data analysis and curve fits were performed using open-source software psignifit toolbox version 2.5.6 for MATLAB version 5 and up (Wichmann & Hill, 2001 a,b).

https://uni-tuebingen.de/en/fakultaeten/mathematisch-naturwissenschaftliche-fakultaet/fachbereiche/informatik/lehrstuehle/neuronale-informationsverarbeitung/research/software/psignifit/

MATLAB Download: https://github.com/wichmann-lab/psignifit/archive/master.zip

python Download: https://github.com/wichmann-lab/python-psignifit/archive/master.zip

All other analyses were performed with custom MATLAB programs. All provided data and code can be accessed with MATLAB. The dataset is structured as follows:

GEVI_data: Each folder corresponds to an animal (e.g. A_ID01, A_ID05, etc.). Within each animal folder, there are multiple folders for different stimulus conditions (e.g. “STIM1”, “STIM2”, etc. For lesions, e.g. “A_ID14_X_STIM1”). Within each stimulus condition are multiple days/sessions of imaging (e.g. Awake_04_03_s1.mat, etc.). These files can be loaded and accessed via MATLAB (see “code_files”). Note, due to exclusion of mice that did not express GEVI, animal ID’s in the manuscript deviate from ID’s from the original experiments. The updated ID’s are described in a separate document, “Animal_IDs.xlsx”.

Behav_data: These files contain processed psychometric data from each animal (for basic learning, e.g. “mouse05_learning.mat”. For different stimulus distributions, e.g. “mouse05_range.mat”. For lesions, e.g. “mouse14_lesion.mat”). In addition, this folder contains group data/analysis/modeling results from multiple animals (e.g. “M01_M05_M06_M07_M13_M14_Full.mat”). All files can be loaded and accessed via MATLAB (see “code_files”).

code_files: This folder contains MATLAB programs for all analysis. Note, Psychometric data analysis and curve fits were performed using open-source software psignifit (see above link for download). The sub-folder “Code for figures” contains all necessary MATLAB code to re-produce the main figures of the manuscript. Within each code, loading files, data analysis/statistics and plotting procedures are described step-by-step.

Histology: This folder contains histology images (TIF files) that were used for the manuscript (Fig. 1 and Fig. 6). All other histology images are available from the corresponding author upon reasonable request.

For detailed questions about the data and computer code, please contact the corresponding author garrett.stanley@bme.gatech.edu