Single-cell analysis of transcription kinetics across the cell cycle

ACCEPTED MANUSCRIPT Single-cell analysis of transcription kinetics across the cell cycle Samuel O Skinner, Heng Xu, Sonal Nagarkar-Jaiswal, Pablo R Freire, Thomas P Zwaka, Ido Golding DOI: http://dx.doi.org/10.7554/eLife.12175 Cite as: eLife 2016;10.7554/eLife.12175 Received: 9 October 2015 Accepted: 28 January 2016 Published: 29 January 2016 This PDF is the version of the article that was accepted for publication after peer review. Fully formatted HTML, PDF, and XML versions will be made available after technical processing, editing, and proofing. Stay current on the latest in life science and biomedical research from eLife. Sign up for alerts at elife.elifesciences.org 1 SINGLE-CELL ANALYSIS OF TRANSCRIPTION KINETICS 2 ACROSS THE CELL CYCLE 3 4 Samuel O. Skinner1,2,3, Heng Xu1,2,4, Sonal Nagarkar Jaiswal5, Pablo R. Freire6, 5 Thomas P. Zwaka5,7, Ido Golding1,2,3,4 6 7 1 8 Baylor College of Medicine, Houston, TX 77030, USA 9 2 Center for Theoretical Biological Physics, Rice University, Houston, TX 77005, USA 10 3 Department of Physics, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 11 4 Center for the Physics of Living Cells 12 University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 13 5 Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, TX 77030, USA 14 6 Department of Molecular and Cellular Biology 15 Baylor College of Medicine, Houston, TX 77030, USA 16 7 17 Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA Verna and Marrs McLean Department of Biochemistry and Molecular Biology Department for Developmental and Regenerative Biology 18 19 Contact: Ido Golding, , 20 1 21 Abstract: 22 Transcription is a highly stochastic process. To infer transcription kinetics for a gene-of-interest, 23 researchers commonly compare the distribution of mRNA copy-number to the prediction of a 24 theoretical model. However, the reliability of this procedure is limited because the measured 25 mRNA numbers represent integration over the mRNA lifetime, contribution from multiple gene 26 copies, and mixing of cells from different cell-cycle phases. We address these limitations by 27 simultaneously quantifying nascent and mature mRNA in individual cells, and incorporating cell- 28 cycle effects in the analysis of mRNA statistics. We demonstrate our approach on Oct4 and 29 Nanog in mouse embryonic stem cells. Both genes follow similar two-state kinetics. However, 30 Nanog exhibits slower ON/OFF switching, resulting in increased cell-to-cell variability in mRNA 31 levels. Early in the cell cycle, the two copies of each gene exhibit independent activity. After 32 gene replication, the probability of each gene copy to be active diminishes, resulting in dosage 33 compensation. 34 35 36 2 37 Introduction: 38 Gene expression is a stochastic process, consisting of a cascade of single-molecule events 39 (Coulon et al., 2014; Sanchez and Golding, 2013), which get amplified to the cellular level. A 40 dramatic consequence of stochastic gene expression is that individual cells within a seemingly 41 homogenous population often exhibit significant differences in the expression level of a given 42 gene (Raj and van Oudenaarden, 2008). In fact, cell-to-cell variability in expression levels is the 43 most commonly used proxy for the presence and magnitude of stochastic effects (Elowitz et al., 44 2002; Raj et al., 2006; Raser and O'Shea, 2005). The mapping between stochastic kinetics and 45 population heterogeneity can be made rigorous by making specific assumptions about the 46 kinetics of gene activity and using stochastic theoretical modeling to predict the copy-number 47 statistics of mRNA or protein that would result from these kinetics (Friedman et al., 2006; Raj et 48 al., 2006; Shahrezaei and Swain, 2008; Thattai and van Oudenaarden, 2001). The theoretical 49 prediction is then compared to measured single-cell data, to validate the assumptions and 50 estimate kinetic parameters. Using this approach, cell-cell variability in mRNA numbers has 51 been successfully used to demonstrate the bursty, non-Poissonian nature of mRNA production 52 in organisms from bacteria to mammals (Bahar Halpern et al., 2015b; Raj et al., 2006; Senecal 53 et al., 2014; So et al., 2011; Zenklusen et al., 2008), and to decipher how genetic and cellular 54 parameters modulate these kinetics (Jones et al., 2014; Sanchez and Golding, 2013). 55 However, the ability to map back mRNA copy-number statistics to transcription kinetics is limited 56 by a number of factors. First, the measured number of mRNA molecules in the cell represents 57 temporal integration over the lifetime of mRNA molecules (Raj et al., 2006). And while in 58 bacteria this lifetime is very short (~mins (Chen et al., 2015)), in higher organisms it can be as 59 long as hours (Schwanhausser et al., 2011). Consequently, the measured mRNA level is a poor 60 proxy for the instantaneous activity of the gene. Second, the cellular mRNA combines 61 contributions from all copies of the gene of interest—for example, four copies in a diploid cell at 3 62 G2. Each of these gene copies acts individually and stochastically (Hansen and van 63 Oudenaarden, 2013; Levesque et al., 2013); their combined contribution depends on whether 64 they are correlated and how. Finally, the sampled population typically contains a mixture of cells 65 at different phases of the cell cycle. As a result, deterministic changes in gene copy number and 66 activity along the cell cycle add to the measured population heterogeneity, and may be 67 erroneously interpreted as resulting from stochastic effects (Zopf et al., 2013). 68 Here we demonstrate how these limitations can be overcome, such that mRNA statistics is 69 reliably used to infer the kinetic parameters of stochastic gene activity. Specifically, we 70 investigate the transcriptional activity of Oct4 and Nanog, two key genes in the pluripotency 71 network of mouse embryonic stem cells (Young, 2011). Elucidating the stochastic kinetics of 72 these genes, and how it changes along the cell cycle, is crucial for understanding pluripotency 73 and the path to differentiation. For one, Nanog expression has been reported to exhibit large 74 cell-to-cell variability (Filipczyk et al., 2013; Kalmar et al., 2009; Singer et al., 2014), and this 75 variability was argued to play an important role in differentiation (Abranches et al., 2014; 76 Chambers et al., 2007; Silva et al., 2009), but both the sources and consequences of Nanog 77 variability are still unclear (Cahan and Daley, 2013; Torres-Padilla and Chambers, 2014). It has 78 also been shown that human stem cells’ propensity to differentiate varies significantly between 79 different phases of the cell cycle (Gonzales et al., 2015; Pauklin and Vallier, 2013; Singh et al., 80 (...truncated)