Exploration of Classic Confounders in Lymphoblastoid Cell Lines used to Study Select Antineoplastic Agents

Abstract

OBJECTIVES: Therapeutic response to chemotherapeutic agents in vitro can be studied using immortalized lymphoblastoid cell lines (LCLs). While LCLs provide a valuable model to study heritable factors and anticancer drug reponse in large populations, the results may be confounded by properties inherent to the model. This study is used to explore possible confounders in Choy et al.’s publicially available dataset (Accession#: GSE11582). METHODS: This study utilized Affymetrics U133A array gene expression and phenotypic data for 162 unrelated LCLs. SPSS was used for two-tailed bivariate Pearson correlation analysis comparing relative 6-mercaptopurine (6-MP), 5-fluorouracil (5-FU), and methotrexate sensitivities, growth rate and ATP levels. GeneSpring was used to compare the top and bottom quartiles of relative ATP levels using the unpaired T-test with a significance threshold of 0.001 and Benjamin-Hochberg FDR (n=82). RESULTS: It was found that relative sensitivities of 5-FU and 6-MP are significantly correlated (r2= 0.627, p<0.0001). Furthermore, it was determined that 5-FU sensitivity and growth rate and ATP levels are also correlated; however, no significant correlation was found between growth rate and ATP levles (r2=0.127, p=0.107). Relative ATP level was found to be a more significant determinant of 5-FU sensitivity than growth rate. GeneSpring analysis showed that 1500 genes are differentially regulated based on ATP levels. The gene ontology related to nucleic acid metabolism was overrepresented (p=1.425E-15). CONCLUSIONS: The results above suggest that growth rate and, to a greater extent, baseline ATP levels influence genetic expression of LCLs and may confound in vitro studies of antineoplastic agents.