Experience-dependent modulation of extracellular matrix integrity supports perceptual skill learning and memory
Data files
Jul 14, 2026 version files 13.74 GB
-
Behavior.zip
18.59 MB
-
Images.zip
13.72 GB
-
README.md
11.06 KB
-
Spreadsheets.zip
1.79 MB
Abstract
Perceptual learning refines sensory abilities but requires extensive training, limiting its real-world impact. Understanding its neural mechanisms could accelerate skill acquisition. The extracellular matrix (ECM) has been proposed to regulate learning by degrading to enable synaptic plasticity, and reaccumulating over several days to stabilize the changes. However, this time course does not align with the temporal dynamics of many forms of learning, including perceptual learning, which involves daily performance gains that consolidate between training sessions. To resolve this discrepancy, we tracked and manipulated auditory cortical ECM integrity in Mongolian gerbils during perceptual learning. We found that the ECM undergoes rapid training-induced changes, degrading and returning to baseline within 24 hours of each session. The magnitude of this cycle diminished with continued training, and the cycle ultimately disappeared as performance plateaued. Enzymatic digestion of the ECM with chondroitinase ABC (chABC) impaired perceptual learning, and post-learning chABC treatment destabilized the acquired skill memory. These findings identify the ECM as a key regulator of perceptual learning and support a novel framework in which an experience-dependent decrease in ECM degradation constrains further plasticity, preserving learned representations against interference from new input.
Dataset DOI: 10.5061/dryad.t4b8gtjht
Files and variables
The dataset includes three zip files:
- Behavior.zip contains behavioral data files (MATLAB format)
- Spreadsheets.zip contains quantitative data tables and metadata for each animal (CSV/XLSX)
- Images.zip contains raw histological imaging data (CZI or TIFF format)
Folder Structure
/Behavior/ Adaptive_Training/ [subject name].mat
/Behavior/ Instrumental_Training/ [subject name].mat
/Images/Ablation/[subject name]/*.czi or *ome.tiff
/Images/IHC/[subject name]/*.czi or *ome.tiff
/Images/chABC Pilot/[subject name]/*.czi or *ome.tiff
/Spreadsheets/WFA_PlotProfiles/ A1/[subject name]/ *.csv
/Spreadsheets/WFA_PlotProfiles/ S1/[subject name]/ *.csv
/Spreadsheets/PNNs A1/[subject name]/ *.csv
/Spreadsheets/PNNs/ S1/[subject name]/ *.csv
/Spreadsheets/Other/*.xlsx or *csv
Behavioral Data (MATLAB)
Location: /Behavior/
- Files are provided in .mat format.
- Each file corresponds to one animal.
- Data are organized into two primary structure arrays:
Session is a 1 × N structure array, where each element corresponds to a single behavioral session.
Session(i).Info contains metadata describing the subject and session
Session(i).Data contains trial-by-trial parameters and behavioral data for that session.
Each field in Session(i).Data is a vector of length T, where T = number of trials in that session.
Output is a 1 × N structure array aligned with Session, such that:
Output(i) corresponds to Session(i)
Each element contains summary metrics and fitted model parameters derived from the corresponding session.
Spreadsheet Data
In all spreadsheets, experimental group abbreviations are as follows
- ST-0h: Completed instrumental training and perfused 0 h following final training session
- ST-4h: Completed instrumental training and perfused 4 h following final training session
- ST-24h: Completed instrumental training and perfused 24 h following final training session
- ST-Unpaired: Pseudotrained animals
- PL-2d: Completed two days of adaptive training and perfused 4 h following final training session
- PL-7d: Completed seven days of adaptive training and perfused 4 h following final training session
SubjectSummary.xlsx
Location: /Spreadsheets/Other/
This file provides a centralized summary of metadata for all animals included in the dataset. It serves as the primary reference for linking behavioral, histological, and quantitative data across files and experimental conditions.
Sheet Descriptions
Contains metadata for animals used for mapping of WFA intensities and PNN quantification across different training time points. Each row corresponds to a single animal. Column descriptions can be found below. Data from these animals are represented in Figs. 1-2, Figs. S1-S7, and Figs. S12-13.
Contains metadata for animals administered chABC, saline, or penicillinase prior to adaptive training (aka, perceptual learning, PL). Each row corresponds to a single animal. Column descriptions can be found below. Data from these animals are represented in Fig. 3, and Fig. S9)
Contains metadata for animals administered chABC or saline after adaptive training (aka, perceptual learning, PL). Each row corresponds to a single animal. Column descriptions can be found below. Data from these animals are represented in Fig. 4, and Fig. S9)
Contains metadata for animals administered chABC, saline, or penicillinase prior to instrumental training (aka, instrumental learning, IL). Each row corresponds to a single animal. Column descriptions can be found below. Data from these animals are represented in Fig. S10)
Contains metadata for animals used to test chABC efficacy and time course of extracellular matrix reconstitution. Each row corresponds to a single animal. Column descriptions can be found below. Data from these animals are represented in Fig. S8.
Column Descriptions
Unique identifier for each animal. This ID is used to link data across behavioral files (.mat), spreadsheet data, and imaging files.
Experimental condition assigned to the animal. Groups correspond to training conditions used in the manuscript (e.g., Untrained, PL-2d, PL-7d, etc.), or to timing of perfusion following a chABC pilot (e.g., Injected with chABC and perfused 15 days later).
Histological processing batch. This variable was used to account for batch effects in staining intensity or imaging.
Immunohistochemistry protocol used for tissue processing (e.g., Streptavidin-based labeling).
Biological sex of the animal. M = male. F = female.
Age of the animal at the time of perfusion (in postnatal days).
Time of day at which perfusion was performed.
Number of training days required for the animal to reach behavioral criterion. Values are provided only for trained animals.
Compound infused into auditory cortex. chABC = chondroitinase ABC.
Percentage of auditory cortex with WFA expression.
Percentage of somatosensory cortex with WFA expression.
NumAMTrialsPresented.csv
Location: /Spreadsheets/Other/
-Each row contains data for one animal
-Session columns contain the number of AM trials presented in each session
-Other columns are named using the conventions described above.
NeuNCellCounts.xlsx
Location: /Spreadsheets/Other/
-Each row contains data for one animal
-Area column contains anatomical information using the format: [SLICE]_[HEMISPHERE].
-Example: 1-H1 = Slice 1 (most rostral), hemisphere 1.
-Each ROI column contains data for a region of interest (ROI). Values are NeuN+ cell counts in that ROI.
-Other columns are named using the conventions described above.
NumShocksPresented.xlsx
Location: /Spreadsheets/Other/
-Each row contains data for one animal
-Each session column contains the number of shocks presented during that session.
-Other columns are named using the conventions described above.
Plot Profiles (WFA Intensity Across Cortical Depth)
Location: /Spreadsheets/WFA_PlotProfiles/
- Each file corresponds to one animal.
- Files are organized by cortical region (auditory cortex, A1, or somatosensory cortex, S1)
- Columns are ordered by slice and hemisphere.
- Rows correspond to positions along the cortical depth (sampling axis).
- Values are raw WFA fluorescence intensity (a.u.).
PNN Quantification (WFA Intensity for individual perineuronal nets (PNNs))
Location: /Spreadsheets/PNNs/
- Each file corresponds to one animal.
- Files are organized by cortical region (auditory cortex, A1, or somatosensory cortex, S1)
- Each row represents one PNN (perineuronal net).
Column Descriptions
Anatomical information using the format: [HEMISPHERE]_[SLICE] _[LAYER].
Example: LEFT_1_L2 = Left hemisphere, slice 1, cortical layer 2.
Slices are ordered from rostral to caudal (1 = most rostral, 3 = most caudal)
PNN_TRUE
Indicates which putative PNNs met the threshold for inclusion. Those with a 1 are included; those with a NaN are discarded in the subsequent analysis pipelines.
PV+PNN
Indicates which PNNs were associated with a PV+ cell. (1 = PV+, NaN = PV-).
Mean
Mean WFA intensity of the PNN.
Imaging Data (CZI and TIFF Files)
Location: /Imaging/
- Raw microscopy data in Zeiss .czi or exported to ome.tiff format.
- Organized by experiment (ablation, IHC, or chABC pilot).
- Each file corresponds to a single slice and hemisphere and contains associated image acquisition metadata.
Code/software
Software and Workflow Description
MATLAB code for processing behavioral data files is available at https://github.com/caraslab. MATLAB code for data analysis and figure generation is available at https://github.com/caraslab/Winne_2026/.
Software Requirements
Analyses were performed in:
- MATLAB (R2019b or newer)
- Statistics and Machine Learning Toolbox
To access and reanalyze raw imaging data, additional software is required:
- Fiji/ImageJ (recommended; with Bio-Formats plugin)
- Alternatively: Zeiss ZEN software or other Bio-Formats–compatible image analysis tools
These programs are required to open and process .czi and OME-TIFF image files.
Relationship Between Data and Code
The Dryad dataset and GitHub repositories are designed to be used together:
- Dryad repository: contains all raw and processed data (behavioral .mat files, spreadsheet data, and raw .czi imaging files)
- GitHub repository: contains MATLAB scripts used to process, analyze, and visualize these data
All reported quantitative results and figures can be reproduced by applying the GitHub analysis pipelines to the data provided in Dryad.
Raw imaging files are provided to enable independent reanalysis outside of the MATLAB pipelines.
Analysis Workflow
The analysis is organized into two primary pipelines:
Processes MATLAB behavioral data files (.mat) containing Session (trial-level data) and output (session-level summaries).
Key steps include:
- Extraction of behavioral performance metrics (e.g., thresholds, false alarm rates, trial counts)
- Computation of learning-related measures (e.g., learning rates across training days)
- Aggregation across experimental groups
- Visualization of behavioral performance and psychometric functions
https://github.com/caraslab/Winne_2026/
Processes quantitative spreadsheet data derived from histological images.
Key steps include:
- Aggregation of WFA intensity plot profiles across cortical depth
- Compilation of PNN measurements across animals
- Normalization of intensity values
- Conversion of counts to densities
- Generation of group-level summaries and visualizations
Execution Notes
- File paths in MATLAB scripts must be updated to match local directory structures after downloading the dataset.
- Data files are organized by animal; consistent naming conventions enable linkage across behavioral, imaging, and quantitative datasets.
- All analysis outputs are saved as CSV tables and vectorized PDF figures.
Additional Documentation
Detailed implementation and usage instructions are available in the GitHub repository README. Users interested in reanalyzing raw imaging data should use Fiji/ImageJ or other compatible software to access .czi or OME-TIFF files.