Eribulin induces mesenchymal-epithelial transition to curtail metastatic progression by altering the chromatin landscape of breast cancers cells [scATAC-Seq]

The epithelial-mesenchymal transition (EMT) is a developmental program that is co-opted by tumor cells to aid in their embarking on the metastatic cascade. Tumor cells that undergo EMT are known to be endowed by, amongst other traits, heightened resistance to chemotherapy. Despite a clear understanding of the role played by the EMT program in conferring cells with these aggressive traits, there are currently no therapeutic avenues specifically targeting the program. Here we show that treatment of mesenchymal-like breast cancer cells with eribulin, an FDA-approved drug for the treatment of advanced breast cancers, leads to a mesenchymal-epithelial transition (MET), accompanied by a loss of metastatic propensity and sensitization of tumor cells to subsequent rounds of chemotherapy. We uncover a novel alternate mechanism of action that provides evidence for eribulin pretreatment as a viable clinical mechanism of MET induction that curtails metastatic progression and the evolution of therapy resistance. Mouse cells were sequenced from RNA and nucleus. Eight libraries for single cell/nuclei transcriptome data were obtained in 3 batches namely Run1, Run2 and Run3. Eribulin (ERI) and Paclitaxel (PAC) drugs were considered as treatment with various level of doses. While Run1 and Run2 data were from single cell 3’ V3 sequencing (scRNA) protocol, Run3 was single nuclei RNA+ATAC V1 sequencing (snRNA multiome) protocol. FASTQ and count matrix were generated using 10x Genomics cellranger-3.1.0, 10x Genomics cellranger-4.0.0 and 10x Genomics cellranger-arc-1.0.0 for Run1, Run2 and Run3 respectively. Four libraries for single cell/nuclei ATAC data were obtained in 2 batches namely Run1 and Run3 using scATAC V1 and snATAC (multiome) V1 protocol respectively. FASTQ and peak count matrix were generated using 10x Genomics cellranger-atac-1.2.0 and 10x Genomics cellranger-arc-1.0.0 for Run1 and Run3 respectively. Cellranger reference mm10 was used as reference for all runs and technologies. Count matrices were analysed using Seurat v3.1.4 pipeline (Stuart et al., 2019). Briefly, low quality cells (number of genes expressed < 200, percentage of UMIs in mitochondrial and ribosomal genes individually > 50) are filtered out. Using selected variably expressing genes and significant PCs from PCA, clustering and UMAP projection were generated. Pseudotime analysis was done against all samples merged dataset using Monocle3 v0.2.1 while considering “Untreated” cells as the root cells. Peak count matrices were analysed using Signac v1.1.1 pipeline. Briefly, low quality cells (fragments in peak regions < 1000 and >75000, percentage of UMIs in peaks > 20, ratio of UMIs in blacklist regions to that of peak regions > 0.05, nucleosome signal > 10 and TSS enrichment < 2) are filtered out. Using selected top features and 2 to 30 PCs from LSI, clustering and UMAP projection were generated.