RNA-Seq of multiplexed metastasis xenogfrates of barcoded cell line pools

Large-scale metastasis research has been infeasible as the use of in vivo hosts precluded the scale of such experiments. Here we introduce a barcoding strategy coupled with deep sequencing that is capable of determining the metastatic abilities of human cancer cell lines in murine xenografts at scale. We developed a RNA-Seq based readout method and an analysis framework that co-capture the cell lines that metastasize to the target organs and their averaged transcriptomes in different in vivo microenvironment. We applied the approach to a cohort of basal-like breast cell lines and revealed their diverse metastatic patterns, a heterogeneity as seen in patients. Expression analysis revealed organ-specific transcriptional states and differential pathway activities in metastases from one site to another. Overall design: Individual breast cancer cell lines were labeled with unique barcodes, and barcoded cell lines were pooled at equal ratio, and cell line pools were injected into the left-ventricle of NSG mice. Animals were sacrificed 5 weeks after injection, and organs including brain, lung, liver, kidney, bone were harvested. Organs were dissociated to single cell suspensions, subjected to mouse stroma depletion, and sorting for GFP+ cancer cells. Sorted metastatic cells from different organs were then used for low-input RNA-Seq. The pilot pool contains MDAMB231_BREAST, HCC1954_BREAST, BT549_BREAST, CAL851_BREAST The g1 pool contains BT549_BREAST, HDQP1_BREAST, JIMT1_BREAST, MDAMB157_BREAST, MDAMB436_BREAST, HCC1806_BREAST, HMC18_BREAST, MDAMB468_BREAST, DU4475_BREAST. The g2 pool contains CAL851_BREAST, HCC1143_BREAST, HCC70_BREAST, HCC1395_BREAST, HCC1187_BREAST, HCC38_BREAST, HCC1569_BREAST, HCC1937_BREAST, BT20_BREAST, HCC1599_BREAST. Please note that [1] average count data was generated for some replicate samples (e.g. 'pilot.S5962' data column contains averag data over pilot.59, pilot.60, pilot.61, pilot.62) and thus total 144 processed data columns are provided for total 154 samples [2] For mapping and gene counting per se, standard parameters from the software packages were used, and code will be made available upon request as stated in the manuscript.