Skip to main content
NCBI home page
UKPMC Funders Author Manuscripts logoLink to UKPMC Funders Author Manuscripts
. Author manuscript; available in PMC: 2015 Jul 13.
Published in final edited form as: Mol Cell. 2015 Feb 26;58(1):147–156. doi: 10.1016/j.molcel.2015.01.027

Bursty gene expression in the intact mammalian liver

Keren Bahar Halpern 1, Sivan Tanami 1, Shanie Landen 1, Michal Chapal 1, Liran Szlak 1, Anat Hutzler 1, Anna Nizhberg 1, Shalev Itzkovitz 1,*
PMCID: PMC4500162  EMSID: EMS63963  PMID: 25728770

Summary

Bursts of nascent mRNA have been shown to lead to substantial cell-cell variation in unicellular organisms, facilitating diverse responses to environmental challenges. It is unknown whether similar bursts and gene-expression noise occur in mammalian tissues. To address this, we combine single molecule transcript counting with dual-color labeling and quantification of nascent mRNA to characterize promoter states, transcription rates and transcript lifetimes in the intact mouse liver. We find that liver gene expression is highly bursty, with promoters stochastically switching between transcriptionally active and inactive states. Promoters of genes with short mRNA lifetimes are active longer, facilitating rapid response while reducing burst-associated noise. Moreover, polyploid hepatocytes exhibit less noise than diploid hepatocytes, suggesting a possible benefit to liver polyploidy. Thus temporal averaging and liver polyploidy dampen the intrinsic variability associated with transcriptional bursts. Our approach can be used to study transcriptional bursting in diverse mammalian tissues.

Introduction

Gene expression in unicellular organisms and in mammalian cell lines has been shown to be highly bursty (Blake et al., 2003, 2006; Chong et al., 2014; Dar et al., 2012; Bar-Even et al., 2006; Friedman et al., 2006; Golding et al., 2005; Kærn et al., 2005; Newman et al., 2006; Pedraza and Paulsson, 2008; Raj and Oudenaarden, 2008; Suter et al., 2011). Promoters tend to stochastically transition between a closed, transcription-prohibitive state and an open permissive state, generating bursts of nascent transcripts. This bursting phenomenon can generate intrinsic variability, or ‘noise’, in the mRNA content of isogenic cells. In unicellular organisms such variability may constitute a ‘bet-hedging’ strategy, improving the chances that a clonal population adapts to variable conditions (Chalancon et al., 2012; Eldar and Elowitz, 2010).

An open question is whether single-cell variability is an advantage or a disadvantage in tissues that maintain organismal homeostasis. Unlike unicellular organisms, cells residing in homeostatic tissues coordinately function towards a common physiological goal. One may think that gene expression in such systems would be tuned to an optimal set point with minimal variability among cells. Conversely, expression variability could give rise to sub-populations of cells that can rapidly respond to changing environmental stimuli. Such division of labor may be advantageous for metabolic tissues that modulate their function on a rapid time-scale. The extent of bursty transcription and expression variability in mammalian tissues has so far not been explored.

Quantifying intrinsic variability in a tissue is challenging, because tissues are highly heterogeneous and spatially structured. Tissues are often polarized by directional blood flow or morphogenes and thus the physical location of a cell within the tissue is a major extrinsic determinant of gene expression that must be controlled for. The mammalian liver is a prime example of these features. The liver is composed of repeating anatomical units termed lobules, which are polarized by blood flowing from an upstream ‘periportal zone’ to a downstream ‘pericentral zone’ (Hoehme et al., 2010). These zones differ in the levels of oxygen, nutrients and hormones, as well as the expression levels of genes (Jungermann and Keitzmann, 1996; Gebhardt, 1992; Braeuning et al., 2006). An additional source of heterogeneity in the liver is its polyploidy (Celton-Morizur and Desdouets, 2010; Duncan et al., 2010; Pandit et al., 2013). The liver consists of a mixture of hepatocytes with either one or two nuclei, where each nucleus has 2, 4, 8 or 16 copies of each chromosome. Thus the liver is composed of multiple sub-populations distinguished by ploidy and tissue location. Fewer than 0.1% of hepatocytes are cycling at any given time (Duncan 2013), and so the contribution of cell-cycle stage to gene expression variability in the liver can be neglected. Assessing intrinsic variability among hepatocytes must take into account the key sources of heterogeneity in this tissue and focus on expression differences between cells of identical ploidy residing at the same spatial zone.

Several techniques enabled inference of promoter bursting kinetics from fluorescent reporters (Elowitz et al., 2002; Raser and O’Shea, 2004; Suter et al., 2011), or single-molecule time-lapse studies of promoter dynamics (Darzacq et al., 2007; Larson et al., 2011), however these are not suitable for intact tissues. Single molecule fluorescence in-situ hybridization (smFISH) can identify mature as well as nascent transcripts of endogenous genes (Raj et al., 2008; Zenklusen et al., 2008; So et al., 2011; Boettiger and Levine, 2013; Itzkovitz et al., 2011; Little et al., 2013; Senecal et al., 2014) and has the potential of controlling for both cell type and physical locations of cells within a tissue. Here we develop a methodology based on smFISH to quantitatively characterize promoter states and bursting properties of cells in intact mammalian tissues. We find that bursty transcription is the common mode of gene expression in the liver, however tight coordination of transcription and degradation as well as liver polyploidy reduce the burst-associated noise.

Results

Hepatocytes exhibit extensive intrinsic gene expression variability

To assess the intrinsic variability in the expression of liver genes we imaged individual mRNA molecules in mouse liver frozen-sections using smFISH (Itzkovitz et al., 2011; Lyubimova et al., 2013) (Figure 1, Figure S1). We used simultaneous DAPI nuclear staining and Phalloidin membrane staining to assign mRNA dots to individual cells. We developed an in-situ ploidy classification algorithm (Extended Experimental Procedures) that enabled stratifying our single-cell mRNA counts by both tissue zone and ploidy class (Figure S1D-F). Strikingly, hepatocytes of the same ploidy level and at the same lobule zone exhibited highly variable mRNA levels, spanning up to two orders of magnitude (Figure 1C). Cellular mRNA distributions within these apparently uniform populations were much broader than expected based on a non-bursty one-state promoter model. In contrast, they were well fitted by a two-state bursty model (Extended Experimental Procedures, Figure 1C, Figure 2A), indicating that promoter bursting could be at play.

Figure 1.

Figure 1

Open in a new tab

Single molecule measurements of intrinsic variability in the intact mouse liver. (A) Single molecule transcript counting enables controlling for ploidy and spatial location of cells. Red dots are single mRNA molecules of Pck1, blue are DAPI-stained nuclei, green is Phalloidin membrane staining. PP – periportal zone, PC – pericentral zone. Scale bar is 30 um. Image is a maximal projection of 12 optical sections spaced 0.3um apart. (B) Magnified view of the boxed region in panel (A) showing polyploid hepatocytes with one or two nuclei, each with either 2,4 or 8 copies of each chromosome. (C) Hepatocytes of the same ploidy and tissue zone exhibit substantial intrinsic variability in gene expression. Blue bars are the distributions of the numbers of Pck1 mRNA per cell in tetraploid hepatocytes in the pericentral zone. Red - theoretical probability distribution function (PDF) expected from a one-state non-bursty model, green - distribution of a bursty two-state model. See also Figure S1.

Figure 2.

Figure 2

Open in a new tab

Single molecule detection and quantification of bursting promoters in the intact mouse liver. (A) Two state bursting model of gene expression. f is the fraction of promoters that are actively transcribing, μ is the transcription rate from an active promoter, δ is the mRNA degradation rate and n is the number of gene copies (ploidy). kON, kOFF are the rates of promoter opening/closing respectively. (B) Dual color labeling of introns and exons reveals active transcription sites (TS). Red dots are Actb mRNA detected using a probe library targeting the exons. Green dots are pre-mRNA detected using a probe library targeting the introns. (C) Actb exhibits rare TS with low transcription rate and stable mRNA (low f, μ and δ). (D) Pck1 exhibits abundant intense TS and high degradation rates (high f, μ and δ). Outlined are two adjacent hepatocytes with substantial difference in transcript counts. Arrowheads mark TS. (E) Ratio of intensities of exon dots at TS to those of mature mRNA facilitates extracting polymerase occupancy and transcription rate (μ). Shown are the distributions of the exonic channel intensities of Pck1 for non-TS dots (top) and TS dots (bottom). Inset shows dot examples. Gray ovals represent Pol2 molecules, green dots represent smFISH probes. See also Figures S2 and S3.

Dual-color labeling of introns and exons facilitates quantification of promoter states, transcription rates and transcript lifetimes in-situ

The single-cell variability we observed in hepatocytes may not necessarily be a result of promoter bursting. There could be additional extrinsic sources of variability other than physical location and ploidy among the studied cell population. To directly link variability to promoter bursting we therefore sought to develop a method that would enable unambiguous identification and quantification of promoter states and transcription rates in-situ (Figure 2A). To this end we designed a second smFISH probe library coupled to a different fluorophore, targeting the intron gene segments. Introns are spliced and degraded co-transcriptionally (Levesque and Raj, 2013; Vargas et al., 2011) and indeed intronic probe libraries yielded dots, which resided only in the nucleus (Figure 2B-D). The number of double-labeled dots was smaller or equal to the expected number of loci, estimated from our ploidy classification. These dots disappeared following Actinomycin-D treatment (Figure S2), demonstrating they are indeed active transcription sites (TS). Thus the fraction of actively transcribing promoters (‘burst fraction’, denoted by f, Fig. 2A) could be readily computed from the ratio of the double-labeled nuclear dots and the expected number of gene loci.

We next calculated the number of nascent mRNA residing at each TS, as the ratio of intensities of the exonic TS dot to the intensity of cytoplasmic dots, representing single mRNA (Figure 2E, we included a correction factor for the physical spread of the library along the gene, Extended Experimental Procedures and Figure S3). Using intensity ratios of probe libraries targeting both ends of the gene we demonstrated that >85% of nascent mRNAs at TS are attached to actively transcribing polymerase molecules (Figure S3). Thus the number of nascent mRNA can be used as a proxy for Polymerase occupancy (M). We next used polymerase occupancy to infer the transcription rate μ, the average rate of mRNA production from an active TS, using the equation: μ = M · ν/L, where L is the length of the gene and v=34 ±11bp/s is the polymerase speed, which we calibrated using Actinomycin-D treatment (Experimental Procedures). Finally, we computed the transcript degradation rate δ as the ratio of summed cellular TS transcription rates and total number of cytoplasmic mRNA. We verified our estimates by tracking mRNA decline following cessation of transcription via Actinomycin-D treatment (Extended Experimental Procedures), yielding estimate errors of 15%. Thus our approach enabled measurement of all gene expression parameters (Figure 2A) - the fraction of time a promoter is actively transcribing (burst fraction, f), the transcription rate from an active promoter state (μ) and the mRNA degradation rates (δ), in addition to the total cellular mRNA levels.

Liver genes are expressed in transcriptional bursts

We applied our measurements to 9 liver genes, the gluconeogenic genes Phosphoenolpyruvate carboxykinase 1 (Pck1) and Glucose-6-phosphatase (G6pc), the lipogenic genes ATP citrate lyase (Acly) and Fatty acid synthase (Fasn), the ammonia detoxification genes Argininosuccinate synthase 1 (Ass1) and Glutamine synthetase (Glul), the transcription factor Sterol regulatory element-binding factor 1 (Srebf1), the Insulin receptor (Insr) as well as a housekeeping gene, Beta-actin (Actb). We measured burst parameters in the pericentral and periportal zones in three metabolic states (fasting, fed and highly-fed, Experimental Procedures). We found that promoters of most genes studied were bursting rather than constantly active (Figure 2B-D, Figure 3). Only a subset of the expected gene loci in each nucleus was transcriptionally active (exhibiting nascent mRNA), however those that were active had polymerase occupancies that were too high to be accounted for by a one-state non-bursty model (Figure 3). A notable exception that exhibited non-bursty transcription was Glul, for which all gene loci were transcriptionally active (Figure S4A).

Figure 3.

Figure 3

Open in a new tab

Distributions of cellular mRNA content and of polymerase occupancies fit a two-state bursty model. Left plots for each gene are the probability distribution functions (PDF) of the number of cytoplasmic mRNA per cell, right plots show the PDF of Pol2 occupancies (M). Red is a 1-state non-bursty model fit; green is a 2-state bursty model fit. Both distributions show a better fit to a 2-state bursty model for all genes (Table S1 provides mean square errors of the model fits). All data is for tetraploid hepatocytes in the periportal zone. See also Figures S4 and S5.

The numbers of active TS per cell were binomially distributed among cells for all genes studied (Figure S4B), indicating independent bursting. Moreover, combinatorial labeling (Levesque and Raj, 2013) of TS for Pck1 and G6pc, two functionally related gluconeogenic genes, revealed non-correlated bursting, further supporting intrinsic variability as the source of the differences in gene expression among hepatocytes that reside in the same tissue zone (Figure S4B-C). We used the independent bursting feature to extend a previous analytical two-state promoter bursting model (Raj et al., 2006) to multiple alleles (Extended Experimental Procedures). This enabled inference of the absolute rates of promoter transitions between an ON and OFF state (Figure 2A, Figure 3, Table S1). We find that burst parameters differ widely between genes, ranging from intense frequent bursts for Pck1 (f ≈ 0.73, M≈17 Pol2/TS, kON ≈ 0.65 hr−1, kOFF ≈ 0.24 hr−1) to rare and weaker bursts for Actb (f≈0.03, M≈5 Pol2/TS, kON ≈ 0.08 hr−1 kOFF ≈ 2.2 hr−1 Table S1).

To assess whether the bursty gene expression observed is unique to the liver we repeated our measurements on an additional metabolic tissue – the mouse small intestinal epithelium. We found that most genes were expressed in transcriptional bursts in this tissue as well (Figure S5). Actb was expressed in a non-bursty manner in the intestine, and in a bursty manner in the liver, whereas Glul, which was non-bursty in the liver, exhibited bursty transcription in the intestine (Figure S5).

Burst fraction and transcript degradation rates are tightly correlated to enable rapid response while minimizing noise

Different combinations of burst fraction, transcription rate, and mRNA stability can achieve a required level of cellular mRNA with differential impact on features such as noise and response time (Rabani et al., 2011; Schwanhäusser et al., 2011). For example, coordinated high transcription and degradation facilitates rapid changes in gene expression but can increase intrinsic noise stemming from promoter bursting (Figure 4A). Increasing burst fraction coordinately with degradation rate can dampen the increase in noise. Indeed, we find a tight positive correlation between burst fraction and degradation rate for all genes and conditions studied (R = 0.78, p = 2.2 · 10−5, Figure 4B). Thus genes that should respond fast tend to be less bursty than genes with long mRNA lifetimes. Conversely the long mRNA lifetime we observe for the bursty genes could be a mechanism for minimizing burst-associated noise through temporal averaging of the stochastic burst events.

Figure 4.

Figure 4

Open in a new tab

Burst parameters facilitate rapid response while minimizing noise. (A) Coefficient of variation (C.V.) of cellular mRNA among cells with identical mean expression (set to 100 mRNA/cell) increases with degradation rate and decreases with burst fraction. Supplementary equation [18] was used with to kON = 0.5 to generate the distributions used for calculating C.V. μ was tuned for every combination of f and δ to maintain the same steady state. (B) 3D expression parameter space of liver genes. PP-periportal zone, PC- pericentral zone, f prefix denotes fasting state, h prefix denotes highly-fed state. (C) Projection of gene expression space of G6pc (blue) and Pck1 (red) in the periportal zone in fasting (dashed lines) and high-fed (solid lines) metabolic states. X-axis is production rate (β = 4 · f · μ), Y-axis is degradation rate (δ). Data are represented as mean +/− SEM. (D) Projections of the gene expression space on transcription rate (μ) and burst fraction (f) axes. Higher expression in fasting states is attained by increasing burst fraction and degradation rates and to a lesser extent transcription rate. Data are represented as mean +/− SEM. Lines in (C-D) have identical mean transcript level, obtained by different combinations of the gene expression parameters.

To further explore noise-response tradeoffs in liver gene expression parameter space we performed our measurements on Pck1 and G6pc, the key genes controlling hepatic glucose output, in fed, fasting and re-fed mice. These conditions have been shown to lead to drastic changes in mRNA levels for these genes (Gebhardt, 1992; Jungermann and Keitzmann, 1996). We find that Pck1 and G6pc are up-regulated in fasting conditions through a coordinate increase in both transcript production (β = n · f · μ) and degradation rates (δ) compared to a high-fed state (Figure 4C, D). High degradation rates enable a rapid decline in transcript numbers after a 1-hour period of refeeding (Figure 5). Interestingly the increased transcript production is mainly a result of an increase in burst fraction (f, 10-fold increase for G6pc and 40-fold increase for Pck1, Table S1) and a more modest increase in transcription rate (μ, 1.4-fold for G6pc and 6-fold for Pck1, Figure 4D, Table S1). The increase in transcript production rate predominantly via increased burst fraction is consistent with a strategy of minimizing burst-associated noise (Figure 4A).

Figure 5.

Figure 5

Open in a new tab

Elevated mRNA and protein degradation rates facilitate rapid decline in mRNA and protein levels for G6pc over 1 hour. (A-B) G6pc mRNA (red dots) in mice fasted for 5 hours before (A) and after (B) 1 hour of refeeding. Arrowheads mark TS. Scale bar is 5um. (C) Decrease in transcription rate (left), cytoplasmic mRNA concentration (middle) and protein concentration (right) over an hour of refeeding. (D) Representative western blot for G6PC protein used to calculate the decline in G6PC protein levels presented in (C). α-Tubulin (bottom) was used as a loading control. Data are represented as mean +/− SEM. See also Figure S6.

High mRNA degradation rates of G6pc impact protein dynamics, mRNA intra-cellular localization and correlations of mRNA content with transcription rates

Our in-situ measurements indicated that the gluconeogenic gene G6pc has a particularly short transcript lifetime of ~20-30 minutes (Fig. 4B, Table S1, and Figure 5). To assess the impact of this feature on protein content we measured G6PC protein levels before and after 1-hour of refeeding. Strikingly, we observe not only an almost complete shut-down of transcription and a decline of ~70% in mRNA concentrations, but also a decline of ~60% in protein concentrations over this period (Figure 5C,D). This decline indicates that G6PC protein is also highly unstable under these conditions (protein half-life of 10-45 minutes, depending on whether translation rates change, Extended Experimental Procedures). This half-life is particularly short considering that median protein half-lives are on the order of 50 hours in mammalian cells (Schwanhäusser et al., 2011).

Transcripts of G6pc also exhibited non-random localization of mRNA with the majority of mRNAs spatially clustered around the nucleus (Figure S6A). Considering previous estimates of mRNA diffusion rates of ≈0.03um2/s (Vargas et al., 2005), this clustering could suggest that the mRNA lifetime for this gene may be smaller than the diffusion time within the large hepatocyte volume. At the single-cell level we observed significant correlation between cellular transcription rates and mRNA levels for unstable transcripts such as G6pc, but not for stable transcripts such as Actb (Figure S6B,C). The levels of short-lived mRNA are expected to track the instantaneous promoter activity more tightly compared to long-lived mRNA, supporting our estimates for mRNA lifetimes (Taniguchi et al., 2010). Thus mRNA degradation rates impact the cellular localization of transcripts in hepatocytes, as well as the correlations between instantaneous transcription rates and cellular mRNA levels.

Hepatocyte polyploidy reduces gene expression noise

Unlike most tissues in our body, which consist of mono-nucleated cells with diploid genomes, the liver is a polyploid tissue, consisting of hepatocytes with either one or two nuclei where each nucleus has either 2,4, 8 or 16 copies of each chromosome. This feature is highly ubiquitous with more than ~85% of hepatocytes harboring more than 2 genomic copies in the tissues studied (15% diploids, 75% tetraploids, 8% octoploids). The functional advantages of liver polyploidy remain unclear. Our observation of independent promoter bursting (Figure S4) suggested that polyploidy could serve as an additional mechanism to reduce burst-associated noise.

The noise reduction potential of polyploidy could be understood by considering the distinct effects of polyploidy on averages and variability of mRNA concentrations. A tetraploid hepatocyte has twice the number of copies for each gene compared to a diploid hepatocyte, as well as twice the volume (Pandit et al., 2013), hence the average mRNA concentration should be the same if there is no ploidy-specific regulation. In contrast, variability between the cytoplasmic concentrations among tetraploid cells should be lower than for diploid cells due to the averaging of more stochastic independent events (Figure 6, Extended Experimental Procedures). Indeed, for 8 out of 10 genes and conditions for which median mRNA concentrations were not significantly different between diploids and tetraploids, the coefficients of variation (C.V.) of the concentrations were significantly lower in tetraploid cells compared to diploid cells (Fisher’s combined probability p < 10−16 for all genes tested, median reduction of 13% in C.V.), and no gene had a statistically significant higher C.V. in tetraploids (Figure 6C). These results suggest that polyploidy may be an additional mechanism that could serve to reduce gene expression noise.

Figure 6.

Figure 6

Open in a new tab

Tetraploid hepatocytes have reduced gene-expression noise compared to diploid hepatocytes. (A) Examples of Actb expression among diploid cells (left) and tetraploid cells (right). Cell outlines (white dashed lines) are based on phalloidin membrane staining. Scale bar is 5um. (B) Probability distribution function of mRNA concentrations in diploid (gray) and tetraploid (black) hepatocytes for Actb in the periportal zone of a 5-month old mouse in a fed state. Coefficients of variation (C.V.s) are 0.35 for diploids and 0.28 for tetraploids. (C) Single-cell variability in cytoplasmic mRNA concentrations are smaller in tetraploid (4n) hepatocytes compared to diploid (2n) hepatocytes. Every dot is a gene in one of the conditions studied, shown are the ratios of mean cytoplasmic concentration (blue), and C.V. (green). All genes and conditions analyzed had mean concentrations that were not significantly different, and 8 out of these 10 genes had significantly lower C.V.s. in tetraploids.

Discussion

Our work provides direct evidence of bursty gene expression in intact mammalian tissues. We found that liver promoters stochastically switch between transcriptionally active and inactive states, generating intrinsic variability between cells that are considered identical in terms of ploidy and tissue location. Interestingly the liver seems to possess features that can dampen this variability, through temporal averaging and polyploidy. These two mechanisms effectively increase the number of stochastic transcriptional burst events that contribute to cellular mRNA levels.

We found that mRNA lifetimes of the more bursty genes, the ones that are transcriptionally active for shorter periods, tend to be longer, reducing the burst-associated variability through temporal averaging. As a result of this tight correlation between mRNA degradation rates and burst fractions, gene expression parameter space (the 3-dimensional space spanned by burst fraction, transcription rate and degradation rate) is relatively sparse (Figure 4B). If the liver evolved strategies to reduce noise, what then could be the functional advantage of bursty transcription in this tissue? Non-bursty transcription, consisting of open chromatin continuously transcribing mRNA, would have been more effective than bursty transcription in reducing variability. One possible advantage of the bursty transcription observed could be protection of ‘closed’ DNA from damage, a feature that has been previously attributed to nucleosome structure (Chen et al., 2012). Such protection could be particularly important given the detoxification roles of the liver. An additional advantage of reducing the amount of accessible DNA may be the minimization of miss-binding events of transcription factors (Shinar et al., 2006).

The liver is a polyploid tissue containing either mono-nucleated or bi-nucleated cells where each nucleus contains 2,4,8 or 16 chromosomal copies. The benefits of polyploidy to mammalian cells in general, and particularly for the liver remain unclear (Duncan 2013; Pandit et al., 2013; Storchova and Pellman, 2004; Tang and Amon, 2013). Existing theories include the ability to harbor backup copies of genes to protect cells against mutations. Other advantages could include the higher biosynthetic capacity of polyploid cells, supporting a larger cell size, the ability to generate functional diversity (Duncan et al., 2010) and the ability to modulate surface to volume ratios (Storchova and Pellman, 2004). Our work uncovered an additional possible benefit of polyploidy - reducing gene expression noise. Tetraploid hepatocytes have twice the number of gene copies as well as twice the volume as diploid hepatocytes and indeed we find that the average expression of genes in diploid and tetraploid hepatocytes is mostly unchanged (Figure 6). In contrast to the average expression, variability of mRNA concentrations tends to be lower in tetraploid hepatocytes compared to diploid hepatocytes. This effect can be achieved by spatially averaging more stochastic burst events per cell. Similar spatial averaging effects have been demonstrated in a polyploid mutant of Bacillus subtilis (Süel et al., 2007) and in Drosophila embryos, which are syncytia containing multiple nuclei (Little et al., 2013). Liver polyploidy increases with age (Celton-Morizur and Desdouets, 2010), as does single-cell variability (Bahar et al., 2006). An intriguing hypothesis is that liver polyploidization may counteract the noise increase caused by genes becoming burstier with age. Examining the relation between polyploidy and gene expression noise in liver aging, in liver pathology and in other polyploid tissues such as heart and muscles could provide important insight into the role of this enigmatic feature of mammalian tissues. Measuring the impact of hepatocyte noise on the performance functions of different liver metabolic tasks and comparing metabolic performances in mouse models with perturbed polyploidy (Pandit et al., 2012) can reveal the functional significance of the noise reduction feature of liver polyploidization described in this study.

The liver switches between two distinct modes in terms of glucose metabolism – a glucose absorber following a meal and a glucose producer in between meals. Much of the control of this process is achieved by modulating mRNA levels of the gluconeogenic genes Pck1 and G6pc, the expression levels of which are high in a fasting state and low following a meal. Changes in the expression levels of these genes can be achieved by modulating burst fraction, transcription rate, degradation rate or any combination of these three key features. We found that the higher levels of gluconeogenic genes in a fasting state are achieved via an increase in both transcription and degradation compared to a fed state. The increase in degradation rate comes at a price as it requires excessive energy expenditure for transcription and accentuates burst-associated noise. A possible explanation for the elevated degradation rates during fasting may lie in the fact that the transition from a fasting to a fed state is rapid, as blood glucose levels rise following a meal within minutes. Upon feeding it may thus be important to shut down hepatic glucose output as fast as possible, and indeed for G6pc the high degradation rates facilitate rapid reduction in both mRNA and protein levels following re-feeding (Figure 5). In contrast the transition from a fed to a fasting state takes hours, as metabolites are gradually depleted from the circulation, and thus in a fed state maintaining high mRNA degradation rates (the determinant of response time) may be less important. It will be interesting to explore the mechanisms that facilitate the rapid reduction in mRNA and protein levels during refeeding. Possible regulatory candidates may include microRNA, which have been shown to have important roles in maintaining liver zonation (Sekine et al., 2009).

Our approach for measuring all gene expression parameters, namely burst fractions, transcription rates and degradation rates in single cells within intact tissues opens new avenues for exploring the design principles that may have shaped gene expression parameter space. This approach can reveal the extent of transcriptional bursts and gene expression noise in diverse mammalian tissues.

Experimental Procedures

Mice and tissues

C57bl6 male mice age 5 month were fed normal chow ad libitum, fasting or re-fed for the indicated times. Mice were sacrificed at 6AM (high-fed state), 9AM (fed state) and 12 PM (fasting state, for these mice food was removed at 8AM). In the re-feeding experiment (Figure 5) mice were housed under reverse phase cycle, and fasted for 5 hours starting at 7AM. Mice were then re-fed ad libitum for an hour and sacrificed either before refeeding (2 mice) or after refeeding (2 mice). Mice were anesthetized and tissues were harvested and fixed in 4% paraformaldehyde for 3 hours; incubated overnight with 30% sucrose in 4% paraformaldehyde and then embedded in OCT. 25 μm cryosections were used for hybridization of liver tissue and 12 μm sections were used for intestinal tissue.

Hybridization and imaging

Probe library constructions, hybridization procedures and imaging conditions were previously described (Itzkovitz et al., 2011; Lyubimova et al., 2013). To detect cell borders alexa fluor 488 conjugated phalloidin (Rhenium A12379) was added to the GLOX buffer wash. Portal node was identified morphologically on DAPI images based on bile ductule, central vein was identified using smFISH for Glul performed on serial sections. Hepatocytes within the first three layers of the portal node (up to ~50 um distance) were classified as periportal and hepatocytes within the first four layers of the central vein (~60um distance), but excluding the innermost layer were classified as pericentral. This exclusion stems from distinct expression levels, which we often observed in the innermost layer directly bordering the central vein. All results presented in the paper are for tetraploid (4n) hepatocytes, with the exception of the polyploidy analysis (Figure 6).

Data analysis

Ploidy classification was based on the 3D nuclear dimension reconstruction, based on DAPI images. Validation of the classification was done with an smFISH probe for Xist on livers of female mice (Figure S1E). Transcription sites were identified as dots that appeared in both the exon and intron probe channels. Exon dot intensity was used to infer the number of nascent mRNA at each TS. To calibrate Pol2 speed we imaged genes in NIH 3T3 before and at different time points following Actinomycin-D treatment. Burst fraction and transcription rate were measured before treatment and degradation rates were measured based on the reduction in mRNA levels following treatment, and used to fit a single missing variable - Pol2 speed (Extended Experimental Procedures).

Fitting bursting models to the measured distributions

To fit our measured mRNA distributions (Figure 3) we extended the analytical results of Raj et al. (Raj et al., 2006) to multiple gene copies by convolving the mRNA distributions predicted for a single gene copy. This extension is based on the independent bursting property observed, where the probability of each promoter to be in a transcriptionally active state is independent of the states of other promoters in the cell (Figure S4B). In our fits we used the measured burst parameters (f, μ, δ), leaving a single free fit parameter (kON, since f = kON/(kON + kOFF)). A Poisson distribution was used for the mRNA distributions under the 1-state model. For both the 1-state and 2-state model fits we included a correction for the broadening effect caused by volume subsampling, namely counting mRNA dots only in a partial volume of the cell rather than the entire cell (Extended Experimental Procedures).

Polymerase occupancies for both 1-state and 2-state models were fit with Poisson distributions. Importantly, while for the 2-state model we used the measured mean pol2 occupancy of the active TS as the Poisson parameter ⟨M⟩, in the 1-state model the Poisson parameter ⟨M⟩ was obtained from the burst fraction ⟨M⟩ = −log(1 − f), where f is the fraction of double-labeled dots (TS), since under a 1-state model the fraction of genes where active transcription is not observed is 1 − f = e−<M>. The distributions of polymerase occupancies were convolved with a broadening kernel measured based on the intensities of individual mRNA dots (Extended Experimental Procedures). Mean squared errors were based on cross validations to avoid over-fitting (Extended Experimental Procedures).

Comparing noise properties of different ploidy classes

To quantify variability in cytoplasmic concentrations between different ploidy classes we counted the number of mRNA molecules in 10 consecutive Z-stacks in a cytoplasmic rim surrounding the nucleus with a fixed volume of 500 um3. This ensures that variability in concentration will not stem from variability in segmented volumes.

To compare averages and noise of cytoplasmic mRNA concentrations between diploids and tetraploids we considered only genes and conditions for which at least 15 cells were counted from each ploidy class and for which mean mRNA concentration was not significantly different between the ploidy classes (using Wilcoxon rank-sum tests and False Discovery Rate of 15%). To ensure comparison of C.V.s of sets of equal lengths of diploids and tetraploids we sampled 1000 sets of n cells from the tetraploid class (where n is the number of cells in the diploid class, invariably the smaller class of cells for the tissues studied), recalculated C.V. and computed p-values as the fraction of 1000 resampled tetraploid sets that yielded a higher C.V. than the diploid set. False Discovery Rate of 15% was used to control for multiple hypothesis testing. Fisher’s method was used to obtain a combined p-value.

Supplementary Material

Table S1

Table S1 Burst parameters for the different genes and conditions studied in the liver. M - Polymerase occupancy. μ - Transcription rate. δ - Transcript degradation rate. f - burst fraction. kON, kOFF - Rates of promoter transitions between ON and OFF states. MSE – mean square errors between the experimental distributions and the model fits (averages of 100 cross-validation runs). MSE are shown for both the distributions of cytoplasmic mRNA per cell and the distributions of polymerase occupancies per transcription site. PP – Periportal, PC – Pericentral.

Table S2

Table S2 Sequences of the probe libraries.

01

Figure S1 (Related to Figure 1) Single molecule gene expression measurements in the intact liver control for spatial zonation and polyploidy. (A) Transcript counting of Pck1 in the intact liver lobule. Each red dot is a mRNA molecule of Pck1, blue square highlights the periportal zone (PP), brown square highlights the pericentral zone (PC). (B) Heat map of gene expression in A. (C) calculation of the concentration of mRNA in different layers around the central vein. Boxes highlight the pericentral and and periportal zones. (D) Histograms of nuclear diameters extracted from the maximal DAPI cross-section. (E) smFISH for Xist RNA (red dots). Each dot represents an inactive X-chromosome. Shown is an example of an octoploid hepatocyte, with two tetraploid nuclei, each with 2 inactive X chromosome loci. (F) Representative profiles of DAPI area at different optical sections (Z-stack) for bi-nucleated hepatocytes. Blue and Red are the two nuclei profiles. Left – a tetraploid hepatocyte (2 diploid nuclei). Right – an octoploid hepatocyte (2 tetraploid nuclei).

Figure S2 (Related to Figure 2) Actinomycin-D treatment. (A) Images of NIH-3T3 cells before Actinomycin-D treatment (left column) and at different time points following 3μg/ml Actinomycin-D (middle and right columns). Red dots are P21 introns, green dots are P21 exons. (B) Estimating P21 degradation rate. Shown are the natural logarithms of the number of mRNA dots per cell per stack before and at different time points after Actinomycin-D treatment. The slope of log(P21 expression) vs. time is 0.37 ± 0.12 hr−1. (C) Intronic dots of P21 rapidly disappear following actD treatment. Y axis shows the number of TS per nucleus.

Figure S3 (Related to Figure 2) Split probe libraries demonstrate that most mRNA molecules at transcription sites are attached to actively transcribing polymerases. (A-B) diagrammatic representation of the probe libraries used. (C) Intensities of representative TS coupled to L1-a594 and L3-cy5 at the corresponding channels. Cy5 intensity is higher than a594 due to the fluorophore properties. (D) Intensities of representative TS coupled to L1-a594 and L2-cy5 at the corresponding channels. Cy5 intensity in (D) is lower than in (C), indicating that most Pol2 molecules are spread along the gene and not pausing at the 3′ end. Surface heights in (C-D) are the dot intensities in the optical section of maximal intensity. (E) Ratio of the intensities in the alexa594 channel and the cy5 channel for the L1/L3 experiment (blue), and the L1/L2 experiment (green). Red dots are the intensity ratios of L1/L3 multiplied by the theoretical ratio of equation [15] with F=0.86. (F) Gene lengths and Probe spread correction factors for the genes studied.

Figure S4 (Related to Figure 3) Liver promoters burst in a non-correlated manner. (A) Glutamine Synthetase (Glul) exhibits non-bursty transcription in liver tissues. Shown are 6 representative Z-sections out of a total of 45 imaged stacks, spaced 0.3um apart. Green dots are Xist loci marking inactivated X chromosomes, nuclear red dots are TS of Glul, blue are nuclei marked with DAPI. Dashed lines mark nuclei, solid lines mark cells. Three cells for which the entire nucleus appeared in the entire Z-stack are shown - cell a is a mono-nuclear diploid cell showing 1 Xist locus and 2 TS of Glul, cell b is a bi-nucleated tetraploid cell (two diploid nuclei) in which each nucleus exhibits 1 Xist locus and 2 TS of Glul, cell c is a mononuclear tetraploid cell exhibiting 2 Xist loci and 4 TS of Glul. (B) The numbers of active transcription sites (TS) per cell are binomially distributed. Blue bars are experimental measurements; green is a binomial distribution with parameter f, the mean fraction of active TS per site. Data is for tetraploid hepatocytes in the periportal zone. (C) Combinatorial smFISH demonstrates independent bursting of Pck1 and G6pc. Transcription sites for Pck1 (red arrows) were detected using probes targeting Pck1 introns labeled with tmr and probes targeting Pck1 exons labeled with alexa594. Transcription sites of G6pc (green arrows) where detected using probes targeting G6pc introns labeled with tmr and probes targeting G6pc exons labeled with cy5. Although hybridization was performed with two distinct libraries labeled with the same fluorophore (Pck1 and G6pc introns, left), TS for these genes could clearly be detected and quantified based on the appearance of double-labeled dots. Circles denote the nuclei of two adjacent cells (differentiated based on co-staining with Phalloidin, not shown). The number of TS per nuclear volume and the summed occupancies of all TS divided by nuclear volume were not significantly correlated between G6pc and Pck1 (R=0.29, p=0.17 and R=0.25, p=0.24 respectively). Experiments were done on fasting mice.

Figure S5 (Related to Figure 3) Bursty gene expression in the intestinal epithelium. (A) Most genes in intestinal epithelial cells exhibit rare transcription sites (TS) with high Pol2 occupancies that cannot be explained by a non-bursty transcription model. Red dots – mRNA for Pck1 detected with an exon library, blue – DAPI nuclei. Boxed region is magnified on the right, showing single TS with occupancy of 7 Pol2 molecules (yellow arrowhead) at the exon (top), intron (middle) and DAPI (bottom) channels. Scale bar is 5um. Arrow points at a nucleus of a goblet cell. (B) All genes except Actb fit a 2-state bursty model (green) rather than a 1-state non-bursty model (red). For Actb Pol2 occupancies better fit a 1-state model. (C) Burst parameters for the genes studied in the intestinal epithelium.

Figure S6 (Related to Figure 5) Gluconeogenic genes exhibit high degradation rates generating non-random intra-cellular mRNA localization and correlations with transcription rates. (A) Liver genes with short-lived mRNA are clustered around the nuclear periphery. Green dots are G6pc transcripts (left) or Actb transcripts (right). Bottom plots show the histograms of distances of cytoplasmic mRNA to the nuclear periphery. (B-C) Highly unstable genes exhibit positive correlations between cellular transcription rate and mRNA concentrations. (B) G6pc in pericentral zone in fasting state, RSpearman=0.51, p=0.001 (δ=1.24 hr−1), (C) Actb in periportal zone in fed state, RSpearman=0.10, p=0.37 (δ=0.02 hr−1). Every dot in panels B-C is a cell, x-axis is the summed transcription rate from all active TS divided by the cellular ploidy, y-axis is the cellular concentration of mRNA.

Acknowledgements

We thank Uri Alon, Ron Milo, Eran Segal and Jean Hausser for valuable comments. We thank Hagit Shapiro, Sara Weiss and Yael Kuperman for technical help. S.I. is the incumbent of the Philip Harris and Gerald Ronson Career Development Chair. We acknowledge support from the Henry Chanoch Krenter Institute for Biomedical Imaging and Genomics, The Leir Charitable Foundations, Richard Jakubskind Laboratory of Systems Biology, Cymerman - Jakubskind Prize, The Lord Sieff of Brimpton Memorial Fund, The Human Frontiers Science Program, the I-CORE program and the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013) / ERC grant agreement number 335122.

References

  1. Bahar R, Hartmann CH, Rodriguez KA, Denny AD, Busuttil RA, Dollé MET, Calder RB, Chisholm GB, Pollock BH, Klein CA, et al. Increased cell-to-cell variation in gene expression in ageing mouse heart. Nature. 2006;441:1011–1014. doi: 10.1038/nature04844. [DOI] [PubMed] [Google Scholar]
  2. Blake WJ, KÆrn M, Cantor CR, Collins JJ. Noise in eukaryotic gene expression. Nature. 2003;422:633–637. doi: 10.1038/nature01546. [DOI] [PubMed] [Google Scholar]
  3. Blake WJ, Balázsi G, Kohanski MA, Isaacs FJ, Murphy KF, Kuang Y, Cantor CR, Walt DR, Collins JJ. Phenotypic Consequences of Promoter-Mediated Transcriptional Noise. Mol. Cell. 2006;24:853–865. doi: 10.1016/j.molcel.2006.11.003. [DOI] [PubMed] [Google Scholar]
  4. Boettiger AN, Levine M. Rapid Transcription Fosters Coordinate snail Expression in the Drosophila Embryo. Cell Rep. 2013;3:8–15. doi: 10.1016/j.celrep.2012.12.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Braeuning A, Ittrich C, Köhle C, Hailfinger S, Bonin M, Buchmann A, Schwarz M. Differential gene expression in periportal and perivenous mouse hepatocytes. FEBS J. 2006;273:5051–5061. doi: 10.1111/j.1742-4658.2006.05503.x. [DOI] [PubMed] [Google Scholar]
  6. Celton-Morizur S, Desdouets C. Polyploidization of liver cells. Adv. Exp. Med. Biol. 2010;676:123–135. doi: 10.1007/978-1-4419-6199-0_8. [DOI] [PubMed] [Google Scholar]
  7. Chalancon G, Ravarani CNJ, Balaji S, Martinez-Arias A, Aravind L, Jothi R, Babu MM. Interplay between gene expression noise and regulatory network architecture. Trends Genet. 2012;28:221–232. doi: 10.1016/j.tig.2012.01.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Chen X, Chen Z, Chen H, Su Z, Yang J, Lin F, Shi S, He X. Nucleosomes Suppress Spontaneous Mutations Base-Specifically in Eukaryotes. Science. 2012;335:1235–1238. doi: 10.1126/science.1217580. [DOI] [PubMed] [Google Scholar]
  9. Chong S, Chen C, Ge H, Xie XS. Mechanism of transcriptional bursting in bacteria. Cell. 2014;158:314–326. doi: 10.1016/j.cell.2014.05.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dar RD, Razooky BS, Singh A, Trimeloni TV, McCollum JM, Cox CD, Simpson ML, Weinberger LS. Transcriptional burst frequency and burst size are equally modulated across the human genome. Proc. Natl. Acad. Sci. U. S. A. 2012;109:17454–17459. doi: 10.1073/pnas.1213530109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Darzacq X, Shav-Tal Y, de Turris V, Brody Y, Shenoy SM, Phair RD, Singer RH. In vivo dynamics of RNA polymerase II transcription. Nat. Struct. Mol. Biol. 2007;14:796–806. doi: 10.1038/nsmb1280. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Duncan AW. Aneuploidy, polyploidy and ploidy reversal in the liver. Semin. Cell Dev. Biol. 2013;24:347–356. doi: 10.1016/j.semcdb.2013.01.003. [DOI] [PubMed] [Google Scholar]
  13. Duncan AW, Taylor MH, Hickey RD, Hanlon Newell AE, Lenzi ML, Olson SB, Finegold MJ, Grompe M. The ploidy conveyor of mature hepatocytes as a source of genetic variation. Nature. 2010;467:707–710. doi: 10.1038/nature09414. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Eldar A, Elowitz MB. Functional roles for noise in genetic circuits. Nature. 2010;467:167–173. doi: 10.1038/nature09326. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Elowitz MB, Levine AJ, Siggia ED, Swain PS. Stochastic Gene Expression in a Single Cell. Science. 2002;297:1183–1186. doi: 10.1126/science.1070919. [DOI] [PubMed] [Google Scholar]
  16. Bar-Even A, Paulsson J, Maheshri N, Carmi M, O’Shea E, Pilpel Y, Barkai N. Noise in protein expression scales with natural protein abundance. Nat. Genet. 2006;38:636–643. doi: 10.1038/ng1807. [DOI] [PubMed] [Google Scholar]
  17. Friedman N, Cai L, Xie XS. Linking stochastic dynamics to population distribution: an analytical framework of gene expression. Phys. Rev. Lett. 2006;97:168302. doi: 10.1103/PhysRevLett.97.168302. [DOI] [PubMed] [Google Scholar]
  18. Gebhardt R. Metabolic zonation of the liver: regulation and implications for liver function. Pharmacol. Ther. 1992;53:275–354. doi: 10.1016/0163-7258(92)90055-5. [DOI] [PubMed] [Google Scholar]
  19. Golding I, Paulsson J, Zawilski SM, Cox EC. Real-Time Kinetics of Gene Activity in Individual Bacteria. Cell. 2005;123:1025–1036. doi: 10.1016/j.cell.2005.09.031. [DOI] [PubMed] [Google Scholar]
  20. Hoehme S, Brulport M, Bauer A, Bedawy E, Schormann W, Hermes M, Puppe V, Gebhardt R, Zellmer S, Schwarz M, et al. Prediction and validation of cell alignment along microvessels as order principle to restore tissue architecture in liver regeneration. Proc. Natl. Acad. Sci. 2010 doi: 10.1073/pnas.0909374107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Itzkovitz S, Lyubimova A, Blat I, Maynard M, van Es J, Lees J, Jacks T, Clevers H, van Oudenaarden A. Single molecule transcript counting of stem cell markers in the mouse intestine. Nat. Cell Biol. 2011;14:106–114. doi: 10.1038/ncb2384. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Jungermann K, Keitzmann T. Zonation of Parenchymal and Nonparenchymal Metabolism in Liver. Annu. Rev. Nutr. 1996;16:179–203. doi: 10.1146/annurev.nu.16.070196.001143. [DOI] [PubMed] [Google Scholar]
  23. Kærn M, Elston TC, Blake WJ, Collins JJ. Stochasticity in gene expression: from theories to phenotypes. Nat. Rev. Genet. 2005;6:451–464. doi: 10.1038/nrg1615. [DOI] [PubMed] [Google Scholar]
  24. Larson DR, Zenklusen D, Wu B, Chao JA, Singer RH. Real-Time Observation of Transcription Initiation and Elongation on an Endogenous Yeast Gene. Science. 2011;332:475–478. doi: 10.1126/science.1202142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Levesque MJ, Raj A. Single-chromosome transcriptional profiling reveals chromosomal gene expression regulation. Nat. Methods. 2013;10:246–248. doi: 10.1038/nmeth.2372. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Little SC, Tikhonov M, Gregor T. Precise Developmental Gene Expression Arises from Globally Stochastic Transcriptional Activity. Cell. 2013;154:789–800. doi: 10.1016/j.cell.2013.07.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Lyubimova A, Itzkovitz S, Junker JP, Fan ZP, Wu X, van Oudenaarden A. Single-molecule mRNA detection and counting in mammalian tissue. Nat. Protoc. 2013;8:1743–1758. doi: 10.1038/nprot.2013.109. [DOI] [PubMed] [Google Scholar]
  28. Newman JRS, Ghaemmaghami S, Ihmels J, Breslow DK, Noble M, DeRisi JL, Weissman JS. Single-cell proteomic analysis of S. cerevisiae reveals the architecture of biological noise. Nature. 2006;441:840–846. doi: 10.1038/nature04785. [DOI] [PubMed] [Google Scholar]
  29. Pandit SK, Westendorp B, Nantasanti S, Liere E. van, Tooten PCJ, Cornelissen PWA, Toussaint MJM, Lamers WH, Bruin A. de. E2F8 is essential for polyploidization in mammalian cells. Nat. Cell Biol. 2012;14:1181–1191. doi: 10.1038/ncb2585. [DOI] [PubMed] [Google Scholar]
  30. Pandit SK, Westendorp B, de Bruin A. Physiological significance of polyploidization in mammalian cells. Trends Cell Biol. 2013 doi: 10.1016/j.tcb.2013.06.002. [DOI] [PubMed] [Google Scholar]
  31. Pedraza JM, Paulsson J. Effects of Molecular Memory and Bursting on Fluctuations in Gene Expression. Science. 2008;319:339–343. doi: 10.1126/science.1144331. [DOI] [PubMed] [Google Scholar]
  32. Rabani M, Levin JZ, Fan L, Adiconis X, Raychowdhury R, Garber M, Gnirke A, Nusbaum C, Hacohen N, Friedman N, et al. Metabolic labeling of RNA uncovers principles of RNA production and degradation dynamics in mammalian cells. Nat. Biotechnol. 2011;29:436–442. doi: 10.1038/nbt.1861. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Raj A, Oudenaarden A. van. Nature, Nurture, or Chance: Stochastic Gene Expression and Its Consequences. Cell. 2008;135:216–226. doi: 10.1016/j.cell.2008.09.050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Raj A, Peskin CS, Tranchina D, Vargas DY, Tyagi S. Stochastic mRNA synthesis in mammalian cells. PLoS Biol. 2006;4:e309. doi: 10.1371/journal.pbio.0040309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Raj A, van den Bogaard P, Rifkin SA, van Oudenaarden A, Tyagi S. Imaging individual mRNA molecules using multiple singly labeled probes. Nat. Methods. 2008;5:877–879. doi: 10.1038/nmeth.1253. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Raser JM, O’Shea EK. Control of Stochasticity in Eukaryotic Gene Expression. Science. 2004;304:1811–1814. doi: 10.1126/science.1098641. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Schwanhäusser B, Busse D, Li N, Dittmar G, Schuchhardt J, Wolf J, Chen W, Selbach M. Global quantification of mammalian gene expression control. Nature. 2011;473:337–342. doi: 10.1038/nature10098. [DOI] [PubMed] [Google Scholar]
  38. Sekine S, Ogawa R, Mcmanus MT, Kanai Y, Hebrok M. Dicer is required for proper liver zonation. J. Pathol. 2009;219:365–372. doi: 10.1002/path.2606. [DOI] [PubMed] [Google Scholar]
  39. Senecal A, Munsky B, Proux F, Ly N, Braye FE, Zimmer C, Mueller F, Darzacq X. Transcription Factors Modulate c-Fos Transcriptional Bursts. Cell Rep. 2014;8:75–83. doi: 10.1016/j.celrep.2014.05.053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Shinar G, Dekel E, Tlusty T, Alon U. Rules for biological regulation based on error minimization. Proc. Natl. Acad. Sci. U. S. A. 2006;103:3999–4004. doi: 10.1073/pnas.0506610103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. So L, Ghosh A, Zong C, Sepúlveda LA, Segev R, Golding I. General properties of transcriptional time series in Escherichia coli. Nat. Genet. 2011;43:554–560. doi: 10.1038/ng.821. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Storchova Z, Pellman D. From polyploidy to aneuploidy, genome instability and cancer. Nat. Rev. Mol. Cell Biol. 2004;5:45–54. doi: 10.1038/nrm1276. [DOI] [PubMed] [Google Scholar]
  43. Süel GM, Kulkarni RP, Dworkin J, Garcia-Ojalvo J, Elowitz MB. Tunability and Noise Dependence in Differentiation Dynamics. Science. 2007;315:1716–1719. doi: 10.1126/science.1137455. [DOI] [PubMed] [Google Scholar]
  44. Suter DM, Molina N, Gatfield D, Schneider K, Schibler U, Naef F. Mammalian Genes Are Transcribed with Widely Different Bursting Kinetics. Science. 2011;332:472–474. doi: 10.1126/science.1198817. [DOI] [PubMed] [Google Scholar]
  45. Tang Y-C, Amon A. Gene Copy-Number Alterations: A Cost-Benefit Analysis. Cell. 2013;152:394–405. doi: 10.1016/j.cell.2012.11.043. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Taniguchi Y, Choi PJ, Li G-W, Chen H, Babu M, Hearn J, Emili A, Xie XS. Quantifying E. coli Proteome and Transcriptome with Single-Molecule Sensitivity in Single Cells. Science. 2010;329:533–538. doi: 10.1126/science.1188308. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Vargas DY, Raj A, Marras SAE, Kramer FR, Tyagi S. Mechanism of mRNA transport in the nucleus. Proc. Natl. Acad. Sci. U. S. A. 2005;102:17008–17013. doi: 10.1073/pnas.0505580102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Vargas DY, Shah K, Batish M, Levandoski M, Sinha S, Marras SAE, Schedl P, Tyagi S. Single-Molecule Imaging of Transcriptionally Coupled and Uncoupled Splicing. Cell. 2011;147:1054–1065. doi: 10.1016/j.cell.2011.10.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Zenklusen D, Larson DR, Singer RH. Single-RNA counting reveals alternative modes of gene expression in yeast. Nat. Struct. Mol. Biol. 2008;15:1263–1271. doi: 10.1038/nsmb.1514. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Table S1

Table S1 Burst parameters for the different genes and conditions studied in the liver. M - Polymerase occupancy. μ - Transcription rate. δ - Transcript degradation rate. f - burst fraction. kON, kOFF - Rates of promoter transitions between ON and OFF states. MSE – mean square errors between the experimental distributions and the model fits (averages of 100 cross-validation runs). MSE are shown for both the distributions of cytoplasmic mRNA per cell and the distributions of polymerase occupancies per transcription site. PP – Periportal, PC – Pericentral.

NIHMS63963-supplement-Table_S1.xlsx (17.3KB, xlsx)
Table S2

Table S2 Sequences of the probe libraries.

NIHMS63963-supplement-Table_S2.xlsx (34.5KB, xlsx)
01

Figure S1 (Related to Figure 1) Single molecule gene expression measurements in the intact liver control for spatial zonation and polyploidy. (A) Transcript counting of Pck1 in the intact liver lobule. Each red dot is a mRNA molecule of Pck1, blue square highlights the periportal zone (PP), brown square highlights the pericentral zone (PC). (B) Heat map of gene expression in A. (C) calculation of the concentration of mRNA in different layers around the central vein. Boxes highlight the pericentral and and periportal zones. (D) Histograms of nuclear diameters extracted from the maximal DAPI cross-section. (E) smFISH for Xist RNA (red dots). Each dot represents an inactive X-chromosome. Shown is an example of an octoploid hepatocyte, with two tetraploid nuclei, each with 2 inactive X chromosome loci. (F) Representative profiles of DAPI area at different optical sections (Z-stack) for bi-nucleated hepatocytes. Blue and Red are the two nuclei profiles. Left – a tetraploid hepatocyte (2 diploid nuclei). Right – an octoploid hepatocyte (2 tetraploid nuclei).

Figure S2 (Related to Figure 2) Actinomycin-D treatment. (A) Images of NIH-3T3 cells before Actinomycin-D treatment (left column) and at different time points following 3μg/ml Actinomycin-D (middle and right columns). Red dots are P21 introns, green dots are P21 exons. (B) Estimating P21 degradation rate. Shown are the natural logarithms of the number of mRNA dots per cell per stack before and at different time points after Actinomycin-D treatment. The slope of log(P21 expression) vs. time is 0.37 ± 0.12 hr−1. (C) Intronic dots of P21 rapidly disappear following actD treatment. Y axis shows the number of TS per nucleus.

Figure S3 (Related to Figure 2) Split probe libraries demonstrate that most mRNA molecules at transcription sites are attached to actively transcribing polymerases. (A-B) diagrammatic representation of the probe libraries used. (C) Intensities of representative TS coupled to L1-a594 and L3-cy5 at the corresponding channels. Cy5 intensity is higher than a594 due to the fluorophore properties. (D) Intensities of representative TS coupled to L1-a594 and L2-cy5 at the corresponding channels. Cy5 intensity in (D) is lower than in (C), indicating that most Pol2 molecules are spread along the gene and not pausing at the 3′ end. Surface heights in (C-D) are the dot intensities in the optical section of maximal intensity. (E) Ratio of the intensities in the alexa594 channel and the cy5 channel for the L1/L3 experiment (blue), and the L1/L2 experiment (green). Red dots are the intensity ratios of L1/L3 multiplied by the theoretical ratio of equation [15] with F=0.86. (F) Gene lengths and Probe spread correction factors for the genes studied.

Figure S4 (Related to Figure 3) Liver promoters burst in a non-correlated manner. (A) Glutamine Synthetase (Glul) exhibits non-bursty transcription in liver tissues. Shown are 6 representative Z-sections out of a total of 45 imaged stacks, spaced 0.3um apart. Green dots are Xist loci marking inactivated X chromosomes, nuclear red dots are TS of Glul, blue are nuclei marked with DAPI. Dashed lines mark nuclei, solid lines mark cells. Three cells for which the entire nucleus appeared in the entire Z-stack are shown - cell a is a mono-nuclear diploid cell showing 1 Xist locus and 2 TS of Glul, cell b is a bi-nucleated tetraploid cell (two diploid nuclei) in which each nucleus exhibits 1 Xist locus and 2 TS of Glul, cell c is a mononuclear tetraploid cell exhibiting 2 Xist loci and 4 TS of Glul. (B) The numbers of active transcription sites (TS) per cell are binomially distributed. Blue bars are experimental measurements; green is a binomial distribution with parameter f, the mean fraction of active TS per site. Data is for tetraploid hepatocytes in the periportal zone. (C) Combinatorial smFISH demonstrates independent bursting of Pck1 and G6pc. Transcription sites for Pck1 (red arrows) were detected using probes targeting Pck1 introns labeled with tmr and probes targeting Pck1 exons labeled with alexa594. Transcription sites of G6pc (green arrows) where detected using probes targeting G6pc introns labeled with tmr and probes targeting G6pc exons labeled with cy5. Although hybridization was performed with two distinct libraries labeled with the same fluorophore (Pck1 and G6pc introns, left), TS for these genes could clearly be detected and quantified based on the appearance of double-labeled dots. Circles denote the nuclei of two adjacent cells (differentiated based on co-staining with Phalloidin, not shown). The number of TS per nuclear volume and the summed occupancies of all TS divided by nuclear volume were not significantly correlated between G6pc and Pck1 (R=0.29, p=0.17 and R=0.25, p=0.24 respectively). Experiments were done on fasting mice.

Figure S5 (Related to Figure 3) Bursty gene expression in the intestinal epithelium. (A) Most genes in intestinal epithelial cells exhibit rare transcription sites (TS) with high Pol2 occupancies that cannot be explained by a non-bursty transcription model. Red dots – mRNA for Pck1 detected with an exon library, blue – DAPI nuclei. Boxed region is magnified on the right, showing single TS with occupancy of 7 Pol2 molecules (yellow arrowhead) at the exon (top), intron (middle) and DAPI (bottom) channels. Scale bar is 5um. Arrow points at a nucleus of a goblet cell. (B) All genes except Actb fit a 2-state bursty model (green) rather than a 1-state non-bursty model (red). For Actb Pol2 occupancies better fit a 1-state model. (C) Burst parameters for the genes studied in the intestinal epithelium.

Figure S6 (Related to Figure 5) Gluconeogenic genes exhibit high degradation rates generating non-random intra-cellular mRNA localization and correlations with transcription rates. (A) Liver genes with short-lived mRNA are clustered around the nuclear periphery. Green dots are G6pc transcripts (left) or Actb transcripts (right). Bottom plots show the histograms of distances of cytoplasmic mRNA to the nuclear periphery. (B-C) Highly unstable genes exhibit positive correlations between cellular transcription rate and mRNA concentrations. (B) G6pc in pericentral zone in fasting state, RSpearman=0.51, p=0.001 (δ=1.24 hr−1), (C) Actb in periportal zone in fed state, RSpearman=0.10, p=0.37 (δ=0.02 hr−1). Every dot in panels B-C is a cell, x-axis is the summed transcription rate from all active TS divided by the cellular ploidy, y-axis is the cellular concentration of mRNA.

NIHMS63963-supplement-01.pdf (5.4MB, pdf)

RESOURCES