Clonal differences underlie differential responses to initial, sequential, and prolonged treatment

Introduction: Over the course of a patient’s treatment, they frequently receive multiple drugs. These treatments can either be as combination therapy or sequential treatment after the first has failed. Regardless, rare cancer cells still survive and develop resistance. Our development of complex treatment strategies has outpaced our studies of how such resistance develops in the first place as we continue to study how treatment resistance originates one drug at a time. Instead, we take the stance that studying how rare cells develop resistance to many different treatments could identify cell states to be targeted for broadly reducing treatment resistance. To this end, we performed highly scaled clonal tracing with scRNA-seq to study how clonal populations of melanoma cells survive microenvironmental stress (mimicking hypoxia with CoCl2 or acidic media), small-molecule therapies (MEKi or BRAFi), and broad-spectrum chemotherapeutics (doxorubicin and cisplatin). By tracing clones through initial treatment with each of these agents, we identified cell states that map to resistance to multiple different treatments. We also used this data to ask how clonal populations of melanoma cells change over multiple rounds of treatment and how these changes affect response to secondary treatments. Overall, this work highlights the implications of clonal gene expression states on survival to initial and secondary treatments and suggests that these states could be targeted to decrease treatment resistance.

Materials and

Methods: We used lentivirus to integrate a semi-random DNA barcode library (Emert, et al. 2021) into the genomic DNA of cells. In this system, barcodes are transcribed and polyadenylated. Therefore, these barcodes can be captured using the 10X Genomics 3’ scRNA-seq kit, allowing us to match gene expression with clonal identity for individual cells.

We performed two scRNA-seq and clonal tracking experiments. In Experiment 1, we barcoded melanoma cells and allowed them to go through a limited number of doublings, yielding multiple cells per clone. We then performed scRNA-seq on a subset of cells prior to treatment. We split the remaining cells six ways for four weeks of treatment with either BRAFi, MEKi, acid, CoCl2, doxorubicin, or cisplatin before performing scRNA-seq (Figure 1A). This experiment allowed us to identify clones that had survived each treatment and trace them back to the initial gene expression states.

In Experiment 2, we split the barcoded population into three treatments: combination BRAFi/MEKi, CoCl2, and cisplatin. We measured the transcriptome and clonal identity of cells after a first round of treatment before again splitting the resistant cells three ways for a second round of treatment and scRNA-seq (Figure 1C). This experiment allowed us to identify how different treatment-resistant gene expression states affect a clone's response to further treatment.

For validation experiments, we either flow sorted cells based on resistant marker genes before measuring their survival to treatment over time or performed fluorescent in situ hybridization to measure marker expression over time.

Results, Conclusions, and Discussions: In Experiment 1, we identified CD44 as a recurring marker of clones that would survive different treatments. We validated this computationally-identified marker of being predictive of survival to multiple treatments by flow sorting for CD44-high and CD44-low cells and treating both populations with each of our treatments. We found that CD44-high cells did have improved survival to BRAFi, MEKi, and cisplatin (Figure 1B), showing that our scRNA-seq and barcoding approach identifies gene expression states that predict survival to multiple treatments.

By tracing clones through an additional round of treatment in Experiment 2, we found that treatment resistant clones had variable responses to secondary treatment. We found that surviving initial treatment caused some clones to become resistant to a treatment that they were otherwise sensitive to, termed “induced resistance”. Specifically, we saw that combination BRAFi/MEKi increased expression of IL6ST in clones with induced resistance to CoCl2. We experimentally validated this finding using flow sorting (Figure 1D).

In Experiment 2, we found that clonal differences in gene expression impacted growth during prolonged treatment with the same agent. Clones that developed resistance to combination BRAFi/MEKi were largely either EGFR or NGFR high, but the NGFR-high clones were not detected after the second round of treatment. These results suggest that NGFR-high clones had lower long-term fitness during treatment. We experimentally confirmed that EGFR-high clones were larger than NGFR-high clones after two rounds of treatment using fluorescent in situ hybridization (Figure 1E). The persistence of NGFR-high clones shows that they had maintained their resistance despite having lower long-term fitness.
Overall, these findings emphasize the power of studying the origin of treatment resistance as a complex mapping across multiple different treatments on a clonal basis. We identified a clone-specific gene expression state that broadly marks resistance to multiple different treatments. Further, we show that a clone’s treatment history can alter their susceptibility to future treatment. These findings and methods provide a framework for determining the proper order of treatment to target resistant cell states that emerge over the course of treatment.

Acknowledgements (Optional):