High Content Screen (HCS) for cell spreading of 3T3 fibroblasts.

<p>The major steps of the screen are illustrated: (<b>A</b>) Swiss 3T3 cells are incubated with different compounds in 96-well plates on two different substrates (yellow/red). Each column contains the same compound. The outer wells are left empty. (<b>B</b>) After 1h incubation at 37°C/5% CO₂, cells are fixed with 4% paraformaldehyde in isotonic phosphate buffer and stained with DAPI (nuclei) and Phalloidin (F-actin). (<b>C</b>) 24 images are acquired per channel from each well using the 10x objective. No images are taken in the center or at the border of the well. (<b>D</b>) Image file names are being padded with meta-tags containing location and treatment information. (<b>E</b>) Using a custom developed software pipeline in <b><i>Cell</i></b> <b><i>Profiler</i></b>, cell bodies and cell nuclei are automatically detected. Phenotypic features are extracted and stored in an SQLite database, annotated with its corresponding meta-tags and linked to its compressed image files. (<b>F</b>) The SQLite database can be imported into <b><i>Cell</i></b> <b><i>Analyst</i></b>. Machine learning algorithms can be used to detect the most prominent cell features that are changed in a certain condition. Furthermore, it allows for an automated categorization of cells by their phenotypes (e.g. spread vs. unspread cells). (<b>G</b>) A data analysis pipeline was constructed in <b><i>KNIME</i></b> to automatically import, normalize and visualize the data. Scatter plots/matrices allow for outlier removals and hit selections. Pivot tables of normalized data / hit lists are generated and exported into .csv files for direct import into statistical analysis software. (<b>H</b>) All hits were validated in dose-response assays with 6 instead of 3 replicates.