Skip to main content
Dryad logo

Comparative molecular analysis of cancer behavior cultured in vitro, in vivo, and ex vivo

Citation

Hum, Nicholas R. et al. (2021), Comparative molecular analysis of cancer behavior cultured in vitro, in vivo, and ex vivo, Dryad, Dataset, https://doi.org/10.6071/M3K96Q

Abstract

Current pre-clinical models of cancer fail to recapitulate the cancer cell behavior in primary tumors primarily because of the lack of a deeper understanding of the effects that the microenvironment has on cancer cell phenotype. Transcriptomic profiling of 4T1 murine mammary carcinoma cells from 2D and 3D cultures, subcutaneous or orthotopic allografts (from immunocompetent or immunodeficient mice), as well as ex vivo tumoroids, revealed differences in molecular signatures including altered expression of genes involved in cell cycle progression, cell signaling and extracellular matrix remodeling. The 3D culture platforms had more in vivo-like transcriptional profiles than 2D cultures. In vivo tumors had more cells undergoing epithelial-to-mesenchymal transition (EMT) while in vitro cultures had cells residing primarily in an epithelial or mesenchymal state. Ex vivo tumoroids incorporated aspects of in vivo and in vitro culturing, retaining higher abundance of cells undergoing EMT while shifting cancer cell fate towards a more mesenchymal state. Cellular heterogeneity surveyed by scRNA-seq revealed that ex vivo tumoroids, while rapidly expanding cancer and fibroblast populations, lose a significant proportion of immune components. This study emphasizes the need to improve in vitro culture systems and preserve syngeneic-like tumor composition by maintaining similar EMT heterogeneity as well as inclusion of stromal subpopulations.

Methods

Single-Cell Sequencing and Data Analysis

Tumor growth, digestion, and isolation of cell suspensions were prepared as previously described for tumor digests and tumoroids were dissociated as previously described. Then, 2 subsequent washes in sterile PBS + 0.04% non-acetylated BSA were performed to further remove debris from final suspension. Cell pellets were resuspended in PBS with 0.04% non-acetylated BSA prior to single-cell sequencing preparation using Chromium Single-cell 3ʹ GEM, Library & Gel Bead Kit v3 (10× Genomics, Catalog no. 1000075) on a 10× Genomics Chromium Controller following manufacturers protocol.

Sequencing data was demultiplexed, quality controlled, and analyzed using Cell Ranger (10× Genomics) and Seurat [42]. Data analysis, expression values, and representative plots were generated using Loupe Cell Browser (10× Genomics) and Seurat (26). The Cell Ranger Single-Cell Software Suite was used to perform sample demultiplexing, barcode processing, and single-cell 3′gene counting. Samples were first demultiplexed and then aligned to the mouse genome (mm10) using “cellranger mkfastq” with default parameters. Unique molecular identifier (UMI) counts were generated using “cellranger count”. Further analysis was performed in R using the Seurat package. For in vivo and ex vivo samples, we performed an integrated analysis to identify and compare common cell types. Cells with fewer than 500 detected genes per cell and genes that were expressed by fewer than 5 cells were not included in the analysis. Prior to data integration, we performed a log-normalization and identified the 2000 most variable genes in each dataset. Subsequently, integration anchors were identified and both datasets were integrated to generate a new integrated matrix. The integrated matrix was then scaled to a mean of 0 and variance of 1 and the dimensionality of the data was reduced by principal component analysis (PCA) (30 principle components). Subsequently, a non-linear dimensional reduction was performed via uniform manifold approximation and projection (UMAP) using the first 20 principle components. Then, we used a graph-based clustering approach to cluster cells. We constructed a K-nearest-neighbor (KNN) graph based on the euclidean distance in PCA space using the “FindNeighbors” function and applied Louvain algorithm to iteratively group cells together by “FindClusters” function (resolution = 0.5). A total of 14 clusters were identified in the integrated dataset.

Cell type identification based on high gene expression of the following genes relative to all cells: Cancer: Epcam; Proliferating Cancer: Epcam/Mki67; Fibroblast: Thy1/Dcn; Myofibroblasts: Thy1/Dcn/Acta2; Endothelial: Pecam1/Cdh5; Neutrophils: S100a8/Retnlg; Myeloid: Ptprc/CD14; M2-like Macrophages: Ptprc/CD14/Mrc1/Cd163; Inflammatory macrophages: Ptprc/CD14/Il1b; Proliferating myeloid: Ptprc/CD14/Mki67; T-Cell/NK Cells: Ptprc/Thy1/CD3e/Nkg7; B Cells: Ptprc/CD19/CD79a.

Funding

LDRD, Award: 19-SI-003