Breast cancer (BC) is the most commonly diagnosed cancer among women. Prognosis has improved over the years, to a large extent, owing to personalized therapy informed by molecular profiling of hormone receptors. However, there is a need for new therapeutic approaches for a subgroup of BCs lacking molecular markers, the Triple Negative Breast Cancer (TNBC) subgroup. TNBC is the most aggressive type of BC, lacks an effective standard of care, shows high levels of resistance and relapse is often inevitable. High resistance to therapy has been hypothesized to be associated with high intratumoral phenotypic heterogeneity. To characterize and treat this phenotypic heterogeneity, we optimized a whole-mount staining and image analysis protocol for three-dimensions (3D) spheroids. Applying this protocol to TNBC spheroids identifies the phenotypes of single cancer cells within their spatial context. Cells with dividing, migrating, and high mitochondrial mass phenotypes localize in the outer region of the spheroid. These phenotypes are specifically targeted by Paclitaxel, Trametinib, and Everolimus, respectively, in a dose-dependent manner. Single agents cannot target all phenotypes at the same time, or only at high concentrations which can cause adverse effects. To reduce these adverse effects, we find that targeting multiple cell phenotypes — by combining Trametinib and Everolimus for example — achieves higher cytotoxicity at lower doses. These findings suggest a rational approach to design treatments against heterogeneous tumors and potentially reduce adverse effects.