Bone marrow stromal cells (BMSCs) include a subset of stem cells that are considered promising for developmental studies and therapeutic applications. While it is appreciated generally that BMSC populations can exhibit morphological and functional heterogeneity upon in vitro culture expansion, the potential for heterogeneity within a single colony forming unit–generated ostensibly from a single mother cell–is less explored but is critical to design of both fundamental studies and cell therapy production. Here we observed BMSC colony formation in real time via time lapsed optical imaging and analysis, to quantify whether and how heterogeneity emerged over multiple cell divisions spanning the duration of a typical colony formation unit assay. These analyses demonstrate that such colonies are neither homogeneous subpopulations of stem cells nor necessarily derived from single originating cells. While the mechanisms for and causes of this intracolony heterogeneity are not understood fully, we further demonstrate that extensive cell-cell contacts do not correlate with senescence, but that media exchange was concurrent with diversification in even the most uniform single-cell-derived colonies. These direct quantitative observations and visualizations of colony formation provide new insights that are motivated by significant implications for both basic research and stem cell-based therapies.

Cells sense a wide variety of signals and respond by adopting complex transcriptional states. Most single-cell profiling is carried out today at cellular baseline, blind to cells’ potential spectrum of functional responses. Exploring the space of cellular responses experimentally requires access to a large combinatorial perturbation space. Single-cell genomics coupled with multiplexing techniques provide a useful tool for characterizing cell states across several experimental conditions. However, current multiplexing strategies require programmatic handling of many samples in macroscale arrayed formats, precluding their application in large-scale combinatorial analysis. Here, we introduce StimDrop, a method that combines antibody-based cell barcoding with parallel droplet processing to automatically formulate cell population × stimulus combinations in a microfluidic device. We applied StimDrop to profile the effects of 512 sequential stimulation conditions on human dendritic cells. Our results demonstrate that priming with viral ligands potentiates hyperinflammatory responses to a second stimulus, and show transcriptional signatures consistent with this phenomenon in myeloid cells of patients with severe COVID-19.