Memory and relatedness of transcriptional activity in mammalian cell lineages

Phenotypically identical mammalian cells often display considerable variability in transcript levels of individual genes. How transcriptional activity propagates in cell lineages, and how this varies across genes is poorly understood. Here we combine live-cell imaging of short-lived transcriptional reporters in mouse embryonic stem cells with mathematical modelling to quantify the propagation of transcriptional activity over time and across cell generations in phenotypically homogenous cells. In sister cells we find mean transcriptional activity to be strongly correlated and transcriptional dynamics tend to be synchronous; both features control how quickly transcriptional levels in sister cells diverge in a gene-specific manner. Moreover, mean transcriptional activity is transmitted from mother to daughter cells, leading to multi-generational transcriptional memory and causing inter-family heterogeneity in gene expression.

Quantitative analysis of gene expression and transcriptional memory in mouse embryonic stem cells

Aleksandra Mandic

Quantitative studies of gene expression in have shown widespread variability in transcript and protein product abundance at the single-cell level. Such fluctuations in gene expression have been shown to lead to heterogeneous cellular responses and phenotypes, acting in a range of systems, from immunity, cancer to development. To better understand heterogeneity in cellular populations, we developed a novel method for the precise quantification of gene expression in single living cells. Furthermore, we focused on whether and to which extent transcriptional expression profiles are transmitted during cell division. In the first section, we described a new approach for quantitative measurements of gene expression, based on internal tagging of endogenous proteins with a reporter allowing luminescence and fluorescence time-lapse imaging. Using the currently brightest luminescent reporter, fluctuations of protein expression levels could be monitored in single living cells with high sensitivity and temporal resolution over extended time periods. Moreover, our system allowed measurement of single-cell protein degradation rates in the absence of protein synthesis inhibitors, as well as calculation of absolute synthesis rates over the cell cycle. Finally, the internal tag could be excised by inducible expression of Cre recombinase, which enabled estimation of endogenous mRNA half-lives by monitoring the decay of luminescent signal. Second, to monitor transcriptional activity and quantify memory across multiple endogenous genes in mouse embryonic stem cells, we generated cell lines in which a short-lived luciferase reporter allowed sensitive monitoring of transcriptional kinetics by luminescence imaging at high time resolution over long periods of time. By coupling live single-cell monitoring of transcriptional activity of 9 different promoters with mathematical modelling, we found that different levels of expression fluctuations shape transcriptional memory. Long fluctuations (1-10 generations) and short fluctuations (1-3 h), as well as their amplitudes, vary widely between across different genes and together shape transcriptional memory. Surprisingly, we have shown that there was a large level of gene-specific cell-to-cell synchronicity in transcriptional profiles. Overall, we uncovered that noise and timescales of slow and fast fluctuations, as well as their synchronisation over the cell cycle, defined how transcriptional dynamics were correlated within a lineage of related cells. In conclusion, this thesis described an innovative method that allows measurement of absolute protein expression levels, protein turnover, as well as mRNA half-lives for endogenously expressed genes. Our approach opens new avenues in quantitative studies of gene expression in single living cells. Furthermore, we revealed different levels of transcriptional fluctuations that shape gene specific transcriptional memory and its timescales in mouse embryonic stem cells.

EPFL2018

Conversion of mouse neural stem cells into glioma cancer stem cells

Caroline Turk

Gliomas are the most common and lethal primary adult human brain cancers. The most lethal form of glioma is glioblastoma multiforme (GBM). Multiple apparently differentiated cell types are present within these tumours, including varying proportions of astrocytes and oligodendrocytes. The identification of new GBM subtypes, the role of the cell-of-origin and the role of commonly disrupted genes are active areas of investigation. Of particular interest to this laboratory is the recent realisation that GBMs may be driven by and initiated from tumorigenic stem cells. Our goal was to develop a mouse model of the disease based on genetic manipulation of in vitro neural stem (NS) cell lines. Importantly, this would enable us to define the contribution of each oncogenic pathway in altering the self-renewal of these cells. The original hypothesis was that the simultaneous activation of three critical pathways (RTK/PI3K, Rb and p53) would be required to confer a glioma-like phenotype. However, surprisingly we found that forced expression of the commonly amplified cyclin dependent kinase-4 (CDK4) was sufficient to elicit reduced dependency on growth factor signalling and a decreased sensitivity to the differentiation factor BMP4. Moreover, preliminary results indicate that the mRNA expression of the proto-oncogene Akt2 increases in CDK4 transfectants, suggestive of a feedback mechanism. We tested whether constitutive expression of Akt2 together with CDK4 increases the severity of the phenotype and initial results suggest this is the case. These findings provide encouragement that minimal genetic events will be required in order to transform NS cells in vitro. These cells can now be tested for in vivo tumorigenicity following transplantation. Finally, we have also generated CDK4 overexpressing cell lines in a specific NS cell variant which I identified as expressing an oligodendrocyte differentiation bias (MG1-7). We are therefore now in a position to assess whether differentiation behaviour is affected by the disruption of this pathway. Together these findings highlight the utility of mouse NS cells as a tool to study glioma

2009

Circadian dynamics of RNA localisation in the mammalian liver

Clémence Yumie Syloun Hurni

Gene expression in eukaryotes is a complex multi-step process. It starts in the nucleus with transcription, the synthesis of a mRNA copy from a DNA template. While in the nucleus, the RNA transcript is subject to multiple co- and post-transcriptional modifications, including splicing, capping, polyadenylation, and assembly into ribonucleoprotein complexes. Correctly processed mRNA is exported to the cytoplasm and translated by ribosomes to form a chain of amino acids: the protein. The mRNA lifetime in the cytoplasm is determined by the activity of a distinct set of RNA Binding Proteins, enzymes, and functional RNAs, which promote stability or degradation. Each step of the RNA life cycle is tightly regulated to ensure proper cellular function. The kinetic rates governing these steps are dynamic and notably adapt to the daily fluctuations of the environment related to the alternance of day and night. Indeed, most organisms possess an internal timing system called the circadian clock. The clock is a genetically encoded self-sustained transcriptional-translational feedback loop, which operates in almost every cell and tissue of the body. It controls the temporal gene expression program to synchronise cellular and physiological functions to the external world. In this thesis, I explore the transcriptome of the mouse liver by combining RNA-sequencing, mathematical modelling, and single-molecule RNA-FISH (smFISH), with an emphasis on the spatio-temporal organisation of RNA expression at the subcellular and tissue scales. I first investigate how RNA are differentially localised at the scale of the liver tissue. Hepatocytes are arranged in structural and functional units called lobules, and carry out different physiological functions depending on their spatial position within the lobule. In this collaborative work, we characterised spatio-temporal gene expression profiles, and showed that that while the expression of hundreds of genes is dually orchestrated by time and space, the circadian core clock is expressed uniformly within the liver lobule, and is therefore robust to the heterogeneous microenvironment. Second, I explore the RNA localisation at the scale of a hepatocyte. The subcellular distributions of RNA in different compartments (here, the nucleus and the cytoplasm), are dictated by the balance of a synthesis term and a decay term. To quantify the kinetic parameters driving nuclear and cytoplasmic mRNA accumulation, I sequenced RNA from both cellular fractions from mouse livers sampled at different times of the day. Using a mathematical model describing rhythmic pre-mRNA and mRNA profiles, I could estimate the nuclear export rates and cytoplasmic degradation rates of

\sim

1400 genes. Nuclear export occurs on a much shorter time-scale than cytoplasmic degradation, and nuclear lifetime has only a minor contribution to the total RNA lifetime. However, a subset of metabolic genes remain in the nucleus for more than one hour (up to four hours), which accounts for the long phase delay between the peak times of transcription and of cytoplasmic accumulation. Furthermore, nuclear export contributes to the modulation and generation of rhythmic profiles of ~10% of the cycling nuclear mRNA. This study provides a comprehensive estimation of the nuclear and cytoplasmic life times in the liver and contributes to a better understanding of the dynamic regulation of the transcriptome during the feeding-fasting cycle.

EPFL2021