Pre-existing cell states predict resistance to multiple treatments

Pre-existing differences between individual cancer cells can predict which cells will become resistant to treatment. DNA barcoding methods that track clones and their cell states during treatment have furthered this understanding, previously focusing on resistance to single treatments. Here, we performed multi-treatment, high-throughput clonal tracking and single-cell RNA sequencing to trace rare clones through resistance development across many treatments in parallel, identifying cell states associated with multi-treatment resistance. We found that clones resistant to one treatment had an increased chance of separately developing resistance to other treatments. We identified high CD44 expression in treatment-naive cells as a predictor of future multi-treatment resistance. Additionally, we found that differences in pre-treatment gene expression states can lead cells within the same treatment condition to follow divergent paths toward their ultimate resistance fate. This work provides a framework for extracting targetable gene expression states from complex resistance dynamics to eliminate multi-treatment resistance.