Two-dimensional (2D) cultures are commonly used for testing drug effects largely because of their easy maintenance. But they do not represent the spatial interactions of the cells within a tumor. Three-dimensional (3D) cultures can overcome those limitations thus mimicking the architecture of solid tumor. However, it is not easy to evaluate drug effects in 3D culture for a long time. This necessitates the development of a real-time and longitudinal analysis of 3D platforms. In this study, we transfected the plasmid DNA encoding the fluorescence resonance energy transfer (FRET)-based biosensor into human breast cancer cells and generated two cell lines of MCF7-C3 and MDA-MB-231-C3 (231-C3) cells. We used them to determine the activation of caspase-3, whereby healthy cells appear green and apoptotic cells appear blue by FRET imaging. As the caspase sensors can be constantly produced within the cells and quickly respond to caspase activation, we hypothesized that these sensor cells will allow longitudinal detection of apoptosis. MCF7-C3 and 231-C3 spheroids were generated and subjected to histological examination, gene expression studies, drug treatment, and FRET analyses. Our results demonstrated that MCF7-C3 cells formed tight 3D spheroids, and mimicked in vivo tumor architecture. The mRNA level of tumorigenic markers such as MMP-9, SOX2, and OCT4A were much higher in cells cultured in 3D than in 2D. Finally, upon treatment with paclitaxel, the FRET effect was reduced at the rim of MCF7-C3 spheroids in a dose and time-dependent manner demonstrating these sensor cells can be used to determine drug-induced apoptosis in a 3D set up. This study supports the possibility of developing a biosensor-based in vitro 3D breast tumor model for determination of anti-cancer drug penetration over a long course of time in a non-invasive manner. Biotechnol. Bioeng. 2015;112: 1673-1682. (c) 2015 Wiley Periodicals, Inc.