Purpose: Cell lines are extremely useful tools in breast cancer research. Their key benefits include a high degree of control over experimental variables and reproducibility. However, the advantages must be balanced against the limitations of modelling such a complex disease in vitro. Informed selection of cell line(s) for a given experiment now requires essential knowledge about molecular and phenotypic context in the culture dish.
Methods: We performed multidimensional profiling of 36 widely used breast cancer cell lines that were cultured under standardised conditions. Flow cytometry and digital immunohistochemistry were used to compare the expression of 14 classical breast cancer biomarkers related to intrinsic molecular profiles and differentiation states: EpCAM, CD24, CD49f, CD44, ER, AR, HER2, EGFR, E-cadherin, p53, vimentin, and cytokeratins 5, 8/18 and 19.
Results: This cell-by-cell analysis revealed striking heterogeneity within cultures of individual lines that would be otherwise obscured by analysing cell homogenates, particularly amongst the triple-negative lines. High levels of p53 protein, but not RNA, were associated with somatic mutations (p = 0.008). We also identified new subgroups using the nanoString PanCancer Pathways panel (730 transcripts representing 13 canonical cancer pathways). Unsupervised clustering identified five groups: luminal/HER2, immortalised ('normal'), claudin-low and two basal clusters, distinguished mostly by baseline expression of TGF-beta and PI3-kinase pathway genes.
Conclusion: These features are compared with other published genotype and phenotype information in a user-friendly reference table to help guide selection of the most appropriate models for in vitro and in vivo studies, and as a framework for classifying new patient-derived cancer cell lines and xenografts.
Keywords: Breast cancer cell lines; Digital immunohistochemistry; In vitro model; NanoString.