Pheno-seq – linking visual features and gene expression in 3D cell culture systems

Abstract

Patient-derived 3D cell culture systems are currently advancing cancer research since they potentiate the molecular analysis of tissue-like properties and drug response under well-defined conditions. However, our understanding of the relationship between the heterogeneity of morphological phenotypes and the underlying transcriptome is still limited. To address this issue, we here introduce “pheno-seq” to directly link visual features of 3D cell culture systems with profiling their transcriptome. As prototypic applications breast and colorectal cancer (CRC) spheroids were analyzed by pheno-seq. We identified characteristic gene expression signatures of epithelial-to-mesenchymal transition that are associated with invasive growth behavior of clonal breast cancer spheroids. Furthermore, we linked long-term proliferative capacity in a patient-derived model of CRC to a lowly abundant PROX1-positive cancer stem cell subtype. We anticipate that the ability to integrate transcriptome analysis and morphological patho-phenotypes of cancer cells will provide novel insight on the molecular origins of intratumor heterogeneity.

cola: an R/Bioconductor package for consensus partitioning through a general framework

“…We demonstrate the usage of cola by five gene expression datasets which are the Golub leukemia dataset ( 28 ), the Ritz ALL dataset ( 30 ), the TCGA GBM dataset ( 1 ), the HSMM single-cell dataset ( 14 ) and the MCF10CA single-cell dataset ( 32 ) among which the first three are microarray datasets and the latter two are the RNA-Seq datasets. Four top-value methods (SD, CV, MAD and ATC) and six partitioning methods (hclust, kmeans, skmeans, pam, mclust, and NMF) generating 24 combinations of methods were applied to the datasets.…”

“…Expression values were normalized by log 10 (FPKM + 1) and only the protein-coding genes were used. The MCF10CA single-cell RNA-Seq dataset is available at the EGA database with sample ID SAMEA4666651 and SAMEA4666652 ( 32 ). Raw reads were processed by STAR ( 33 ) and HTSeq-count ( 34 ).…”

Functional States in Tumor-Initiating Cell Differentiation in Human Colorectal Cancer

“…scRNA-seq data suggested that subfractions of MMP low and MMP high spheroid cells preferentially harbor Tdiff-like and Paneth-like (MMP low ) or stem-like and TA-like tumor cells (MMP high ). As spheroid and tumor forming capacity is supposed to be restricted to stem-like tumor cells [ 43 ], we calculated spheroid-forming cell (SFC) frequencies in vitro and TIC frequencies in vivo by limiting dilutions of sorted MMP low and MMP high cell fractions.…”

“…In our proteomic analysis, proteins significantly higher abundant in the MMP high subpopulation included PROX1, one of the top markers of the stem signature and usually expressed in the enteroendocrine lineage [ 51 ]. Interestingly, PROX1 has been reported to be positively correlated with LGR5 expression in CRC [ 43 ] and linked to stem cell maintenance and metastasis [ 68 , 69 ]. Another protein significantly more abundant in MMP high was DEFA6, a protein expressed in normal Paneth and Paneth-like tumor cells [ 70 ].…”

Engineered Tools to Study Intercellular Communication

“…[ 184 ] More recently, patient‐derived cancer cells were cultured as 3D spheroids in barcoded nano‐wells that enabled paired, automated imaging of a 3D culture system with RNA‐seq, termed Pheno‐seq. [ 185 ] Using the Pheno‐seq system, paired imaging and RNA‐seq data of 210 MCF10A spheroids and 95 patient‐derived colorectal cancer spheroids were acquired. Gene expression profiles varied considerably between spheroids that different in size.…”