MEDUSA for Identifying Death Regulatory Genes in Chemo-genetic Profiling Data

Our research program is focused on understanding how different forms of cell death are regulated. We generally study cell death in the context of cancer therapies with the goal of learning how drugs function to turn on cell death and how to make these responses better and more specific. It is becoming increasingly clear that there are many different types of regulated cell death.

The field has currently identified at least 14 mechanistically distinct types of death. Most of these cause a morphologically necrotic death, and in general, we know very little about how these pathways work. An effective way to study cell death regulation is to use functional genomics.

In other words, to systematically perturb each gene and determine how these perturbations affect cell death. When performed in the context of drugs, This is called chemogenetic profiling. Chemogenetic screens typically evaluate how gene knockouts affect the final population size.

However, population size can be influenced by changes in the growth rate, the death rate, or both. Current screening methods fail to identify death regulatory genes due to the confining effects of growth rate variation. Medusa uses simulations of the drug response to model how changing the growth rate or death rate influences the final population size.

These simulations allow us to extract the drug induced death rate for each gene in the screen and remove other confounding factors To begin, plate cells in black, clear-bottom 96 well plates at seeding densities between 1500 and 5, 000 cells per well. Incubate the plates at 37 degrees Celsius with 5%carbon dioxide and humidity for 16 to 24 hours. Prepare drug dilution in media containing the cytotox concentration, optimized previously.

Lyce one plate using Triton X at the optimized concentration. Then place the plate in the plate reader and measure the cytotox fluorescence in the experimental plates at T0 using the optimized settings, and then at a suitable later time point. After the final time point lyce the experimental plates using Triton X to determine the final population size of each condition.

Calculate the fractional viability using the endpoint data and fit the results to dose response curves. Evaluate the dose response curves to identify drug doses that induce approximately 50%cell death. After optimizing the drug dose to induce cell death, randomly assign non-target SG RNAs to non-target genes.

Trim SG RNAs with low counts based on a chosen cutoff strategy. Simulate all possible perturbations to the net population growth rate or NPG by using a for loop to evaluate a range of NPG values. For each experimental single SG RNA, match the simulated fold change closest to the observed data and assign the corresponding relative growth rate to the SG RNA.

Then determine the drug induced growth rate before death from live cell counting data. Fit the data to a simple exponential equation to derive the growth rate. Calculate drug-induced growth rate after death onset by combining the live cell counting data, and the drug grade values.

Plot the grade for the selected drug dose on a grade plot. Using the experimentally determined coordination between drug-induced growth and death rate, create a model to simulate the drug-induced population size over time. Then develop a Medusa model that includes NPG, drug-induced growth rates before and after death onset in doublings per hour, and the drug induced death rate.

Set the starting time point at zero hours and define final time points for both untreated and treated conditions. Simulate all possible perturbations to growth and death rates for the baseline model. For each combination, calculate the logarithm base two of the fold change observed at the assay endpoint.

For each SG RNA, triangulate the drug-induced death rate that would produce the observed fold change. Begin by extracting the predetermined relative growth rate. Determine the rates for drug-induced growth before and after death onset.

Apply the relative growth rate from the NPG to calculate proportional growth rates for drug-induced growth rates before and after death onset. Using the constraints for NPG, drug-induced growth rate before and after death onset, identify the simulated fold change closest to the observed fold change for each SG RNA. Determine gene level growth and death rates for each gene in the library.

Calculate the mean growth and death rates for all SG RNAs associated with each gene, typically ranging from 4 to 10 SG RNAs. To calculate empirical P values, bootstrap the SG RNA level data and apply a false discovery rate correction using the Benjamin Hochberg procedure. Fractional viability of U2 OS cells decreased in a dose dependent manner with approximately 50%lethality observed at five micromolar etoposide after 96 hours.

The recovered population size decreased by 40%to 50%after four days of exposure to five micromolar etoposide, confirming cell death during drug treatment. The grade analysis projected that five micromolar etoposide caused a significant reduction in growth rate with cell death initiated only after complete growth arrest. Medusa analysis revealed that XRCC five knockout cells showed slow growth in untreated conditions and high drug-induced death rates under etoposide treatment.

Validation experiments confirmed that XRCC 5 knockout cells exhibited slower growth in untreated conditions, and elevated death rates under etoposide, aligning with Medusa predictions.