Single Cell Ca2+profiling for the discrimination of cancer cell lines

Abstract

ABSTRACTCalcium (Ca2+) is an essential and ubiquitous second messenger controlling numerous cellular functions. Ca2+signaling relied on the finely tuned oscillations of the cytosolic Ca2+concentrations induced by components of Ca2+signaling toolkit (ion channels, pumps and ion exchangers). The regulation of these Ca2+oscillations define a Ca2+signature that is representative of the cellular identity and phenotype of a cell. In cancers, molecular actors of the Ca2+signaling toolkit are aberrantly expressed. We hypothesized that Ca2+signature of cancer cells are representative of their cellular identity, their tissue of origins (TOO) as well as their isolation site (IS). We defined the Ca2+signature of prostate and colon cancer cell lines by collecting the profile of cytosolic Ca2+responses evoked by a panel of agonists in 22904 individual cells. We then highlighted the heterogeneity of those Ca2+profiles and successfully developed a classifier predicting the tissue of origins (TOO), the isolation site (IS) or the cellular identity of individual cancer cells using a supervised neural network. Unsupervised clustering revealed that Ca2+profiles of single cancer cells derived from 3 main classes of Ca2+responses sub-divided into 50 different clusters. Thus, we highlighted that supervised machine learning applied on top of single cell Ca2+profiling is an effective method to discriminate cancer cells at single cell level and that the cancer cell Ca2+signature can be summarized into 3 main profiles of Ca2+responses.