Fig. 3

Osteosarcoma cells retain phenotypic heterogeneity despite adaptive changes in response to changing microenvironments. A Schematic outlining scRNA-seq bioinformatics workflow. B Osteosarcoma cells maintained overall heterogeneity with a high degree of overlap between conditions. The ridge plot shows ITH scores, which represent the gene expression “distance” between each tumor cell within a sample and all of the other tumor cells from that same sample. The overlap statistic describes the total percentage of overlap in the observed distributions between two samples. C UMAP analysis for merged tibia- and lung-colonizing tumor samples in each of the four models. Cells in gray represent the remaining cells in the merged sample. Cluster enrichment analysis shows the distribution of cells in each cluster in the two microenvironment conditions (tibia, lung). While some cells in the tibia and lung lesions adopted shared phenotypes, others adopted distinct phenotypes. D Metabolic heterogeneity in glycolysis activation. We used a pathway enrichment analysis for hallmark gene sets using genes differentially upregulated in each cluster relative to every other cluster within the same model. P values were adjusted for multiple comparisons. Boxes in gray identify non-significant pathway enrichments, whereas boxes in red identify statistically significant enrichments. Bar plot shows the percentage of cells identified per cluster in lung lesions. In B–D, samples subset to equal number of cells to allow inter-model comparison (n=1500 per condition)