Single Cell Analysis of Population Based Interconversion and Mechanism Study for Cellular Plasticity


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date14 Jan 2019


Increasing evidence suggests that multiple subpopulations of cells with different phenotypes and genotypes coexist within malignant tumors. This tumor heterogeneity has been associated with tumorigenicity, resistance to chemotherapeutic drugs and metastasis. A few models have been proposed to account for the cancer heterogeneity, including clonal evolution, hierarchical models and stochastic model. According to stochastic model, all the cancer cells are equally tumorigenic. Cells can transition from one cell state to another either through intrinsic noise or extrinsic stimuli. Such plasticity plays an important role in cancer progression, such as transition from epithelial state to mesenchymal state during metastasis, dedifferentiation of non-CSCs to CSCs and rapid emergence of multidrug resistance after temporal chemotherapeutic drug exposure. Stochastic interconversion was simulated by Markov model that, under the same microenvironmental condition, transition probabilities from one cell state to another were constant. The transition depended only on cell current state but not from its previously history.

A few studies investigated the interconversion plasticity between two populations of cells. Stochasticity did not always hold true in some of these studies. Some cells remain their original state and showed no sign of interconversion. Additionally, most of these studies only focus on population-based interconversion, rarely were they involved any single cell study. In this study, we attempted to explore the interconversion plasticity at single cell level to better understand the dynamics of phenotypic transition.

To study the dynamic of bulk cell interconversion, we identified and purified two subpopulations, CD90+ and CD90- cells from Lung Adenocarcinoma A549 by Flow Cytometry. Sorted cells were cultured separately and monitored periodically for their CD90 expression. Both CD90+ and CD90- sorted cells slowly transitioned to each other, However, neither one could reconstitute the parental equilibrium, instead each of them stabilize at a new equilibrium that did not converge, indicating a non-stochastic interconversion. We then investigated the interconversion plasticity of each single cell. We divided A549 cells into five fractions based on CD90 expression level and sort individual single cells into 96 well plate. Proliferative clones were further expanded in 96 well plate for further expansion in 6 well plate until reaching 70-80 % confluency to get enough cells to measure clonal CD90 expression in flow cytometry. We revealed that interconversion plasticity of each single cell varies. There was a dependence of clonal CD90 expression on their original expression level. Single cells isolated from CD90+ cells tend to have higher clonal CD90 expression and vice versa. Combining clones of the same fractions obtained an average resemble to the sorted cell equilibrium suggesting a strong correlation of our single cell and bulk cell interconversion.

Among all the clones collected, a portion of them had no sign of interconversion. They maintained stably negative or positive over the course of monitoring. They were defined as committed CD90+ or committed CD90- clones according to their CD90 expression. Clones that harbored both subtypes were defined as transition clones as some cells from the clones must have transitioned from one state to another during the clonal expansion.

We characterized the phenotype of CD90+ and CD90- cells. Both CD90+ from parental cell line and CD90+ committed clones move much faster and is more mesenchymal like than their CD90- counterpart. Interestingly, sorted cells of both types showed no differential tumorigenicity. Conversely, CD90- committed clones were significantly more tumorigenic than CD90+ cells, we therefore reasoned that interconversion blurred the phenotypic difference between two sorted subpopulations.

Gene Set Enrichment Analysis (GSEA) of RNA sequencing data of CD90+ committed cells over CD90- committed cells revealed an upregulation of genes related to “epithelial-mesenchymal transition (EMT)”, “focal adhesion” and “extracellular matrix remodeling” that correlated with higher mobility and mesenchymal characteristic of CD90+ cells. In contrast, genes related to signature oncogenes of lung adenocarcinoma is downregulated, together with a upregulation of genes related to differentiation and development, that correlated with the lower tumorigenicity of the CD90+ cells. In contrary, clinical analysis revealed a significant poor progression free survival for patient with high expression of CD90+ signature genes. Impaired tumorigenicity could be due to 1) expression of genes related to Differentiation/Development or 2) lost plasticity to revert to epithelial state.

We then focused on the factors that regulated the interconversion between CD90+/- and the mechanism for the committed states. We successfully identified a bi-stable switch between HNF1A and CD90. siRNA knockdown of HNF1A successfully promote CD90 mRNA and protein level. Overexpression of CD90 in turn inhibit HNF1A mRNA and protein expression. HNF1A and CD90 therefore mutually inhibit expression of each other. We also identified Collagen I as a putative enhancer of CD90. It promotes CD90 protein expression but not on the mRNA level. TGFβ, an EMT inducing transcription factors significantly increase CD90 mRNA and protein expression. Interestingly, neither Collagen I, TGFβ nor HNF1A could regulate CD90 expression of committed clones.

Committed phenotype was likely due to epigenetic modification. We investigate the enrichment of H3K4me3 and H3K27me3 on CD90 promoter. Only H3K4me3 demonstrates differential enrichment between cells with the highest enrichment of CD90+ committed clones, followed by intermediate parental cell line and the CD90- committed clones. The H3K4me3 enrichment was positively correlated with the CD90 expression therefore it could be one of the key epigenetic regulators for the committed phenotypes. In contrary, H3K27me3 enrichment was negatively correlated with the CD90 expression but the level of enrichment was not significant between different types of cells. DNA methylation did correlate with the CD90 expression. Committed clones of each types have higher enrichment of DNA methylation than intermediate parental cell line. There was insufficient data yet to draw conclusion for how epigenetic regulate cellular plasticity.

In summary, we demonstrated a non-stochastic interconversion by single cells and population-based study. Two subpopulations were identified, migratory mesenchymal like CD90+ cells and epithelial tumorigenic CD90- cells. Interconversion capacity of each single cell varies. Some single cells were able to transition from one cell state to another and generate clones consisted of both CD90+ and CD90-. Interconversion depend on cell original CD90 expression level. Single cells from the extreme end were enriched with clones with no sign of interconversion and were stable over a long period. Bi-stable switch between HNF1A and CD90 may partly contribute to the dependence of clonal CD90 expression to the cells original CD90 state and committed state. The exact mechanism for the committed phenotype is yet to be discovered.