Parabolic growth patterns in 96-well plate cell growth experiments

In preparing for the routine use of the ubiquitous in vitro cell growth inhibition assay for the study of anticancer agents, we characterized the statistical properties of the assay and found some surprising results. Parabolic well-to-well cell growth patterns were discovered, which could profoundly affect the results of routine growth inhibition studies of anticancer and other agents. Four human ovarian cell lines, A2780/WT, A2780/DX5, A2780/DX5B, and A121, and one human ileocecal adenocarcinoma cell line, HCT-8, were seeded into plastic 96-well plates with a 12-channel pipette, without drugs, and grown from 1-5 d. The wells were washed with a plate washer, cells stained with sulforhodamine B (SRB), and dye absorbance measured with a plate reader. Variance models were fit to the data from replicates to determine the nature of the heteroscedastic error structure. Exponential growth models were fit to data to estimate doubling times for each cell line. Polynomial models were fit to data from 10-plate stacks of 96-well plates to explore nonuniformity of cell growth in wells in different regions of the stacks. Each separate step in the assay was examined for precision, patterns, and underlying causes of variation. Differential evaporation of water from wells is likely a major, but not exclusive, contributor to the systematic well-to-well cell growth patterns. Because the fundamental underlying causes of the parabolic growth patterns were not conclusively found, a randomization step for the growth assay was developed.