|Home | About | Journals | Submit | Contact Us | Français|
Mechanisms through which long intergenic noncoding RNAs (ncRNAs) exert regulatory effects on eukaryotic biological processes remain largely elusive. Most studies of these phenomena rely on methods that measure average behaviors in cell populations, lacking resolution to observe the effects of ncRNA transcription on gene expression in a single cell. Here, we combine quantitative single-molecule RNA FISH experiments with yeast genetics and computational modeling to gain mechanistic insights into the regulation of the Saccharomyces cerevisiae protein-coding gene FLO11 by two intergenic ncRNAs, ICR1 and PWR1. Direct detection of FLO11 mRNA and these ncRNAs in thousands of individual cells revealed alternative expression states and provides evidence that ICR1 and PWR1 contribute to FLO11’s variegated transcription, resulting in Flo11-dependent phenotypic heterogeneity in clonal cell populations by modulating recruitment of key transcription factors to the FLO11 promoter.
Two cis-interfering long intergenic ncRNAs, ICR1 and PWR1, regulate transcription of nearby protein-coding gene FLO11 in the yeast Saccharomyces cerevisiae (Bumgarner et al., 2009). These ncRNAs form a bidirectional toggle, one component of a regulatory circuitry that also includes upstream signaling pathways, transcription factors (e.g., activator Flo8 and repressor Sfl1), and chromatin remodelers (e.g., Rpd3L and Hda1 histone deacetylases (HDACs) (Liu et al., 1996; Rupp et al., 1999; Guo et al., 2000; Pan and Heitman, 2002; Halme et al., 2004; Octavio et al., 2009). In their length, position relative to the FLO11 coding region, and effects on FLO11 transcription, ICR1 and PWR1 recall phenomena observed at the yeast SER3 locus (Martens et al., 2004) but are distinct from other types of ncRNA transcription reported at yeast promoters (Seila et al., 2008; Xu et al., 2009; Neil et al., 2009). The ~3.2-kb ICR1 ncRNA initiates ~3.4 kb upstream of the FLO11 ORF and represses FLO11 transcription in cis, whereas ~1.2-kb PWR1 is transcribed from the opposite strand and promotes FLO11 transcription by interfering in cis with ICR1 (Bumgarner et al., 2009). Competitive binding of trans-acting Flo8 or Sfl1 to the FLO11 promoter (Pan and Heitman, 2002) helps to determine which of the two ncRNAs is transcribed (Bumgarner et al., 2009), resulting in alternative FLO11 expression states. Rpd3L− loss-of-function mutants (e.g., cti6) exhibit elevated ICR1 levels, reduced FLO11 expression, and loss of Flo11-dependent phenotypes similar to a flo8 null (Bumgarner et al., 2009). Thus, the HDAC Rpd3L appears to be an activator of FLO11 via repression of ICR1.
The net effect of FLO11’s regulatory circuitry is the variegated transcription of its gene product in clonal wild type (WT) cell populations: FLO11 is expressed (“on”) in some cells and is silenced (“off”) in others (Halme et al., 2004; Bumgarner et al., 2009; Octavio et al., 2009). Expression of Flo11 protein on the yeast cell surface is required for haploid invasion and diploid filamentous growth, which have been understood as foraging responses that occur in nutrient-poor conditions (Roberts and Fink, 1994). Variegated FLO11 expression results in phenotypic heterogeneity within clones because some genetically identical cells differentiate to form filaments that grow away from the founding colony, while others adhere to or invade local surfaces, and still others may wash away to more distant environments (Kaern et al., 2005).
Our previous study of the ncRNA toggle at FLO11 relied on experimental techniques limited in their capacity to capture heterogeneity existing among individual cells in clonal populations. To obtain a more complete view of the roles of ICR1 and PWR1 in regulating FLO11, particularly in view of its variegated expression, we here use fluorescence in situ hybridization (FISH) and fluorescence microscopy to visualize simultaneously coding and non-coding RNA transcripts in fields of intact yeast cells (Raj et al., 2006; Femino et al., 1998; Raj et al., 2008; Zenklusen et al., 2008; Raj and van Oudenaarden, 2008; Pena et al., 2009; Lu and Tsourkas, 2009). These single-cell studies have revealed insights about alternative expression states for FLO11. The data provide evidence at single-cell resolution that ICR1 and PWR1 contribute causally to FLO11’s variegated expression, exerting their effects by modulating the recruitment of key transcription factors (Liu et al., 1996; Pan and Heitman, 2002). Computational modeling combined with single-cell and bulk-cell experimental methods have revealed mechanistic aspects of the regulatory circuitry at FLO11 and may prove useful for investigating roles of ncRNAs across eukaryotic organisms (Guttman et al., 2009; Huarte et al., 2010; Bertone et al., 2004; David et al., 2006; Davis and Ares, 2006; FANTOM Consortium, 2005; van Dijk et al., 2011).
RNA FISH experiments directly demonstrate that FLO11 mRNA variegates in clonal populations of WT yeast (Figure 1B; Figure S1 available online). Previous observations of FLO11 variegation (Halme et al., 2004; Bumgarner et al., 2009; Octavio et al., 2009) relied upon indirect protein-based reporters. Here, we performed quantitative RNA FISH (Raj et al., 2008; Zenklusen et al., 2008) and fluorescence microscopy to detect FLO11 mRNAs at single-cell resolution. Transcripts were imaged in situ in fields of clonal WT cells (Figure 1B). In Z-dimensional image stacks, bound fluorescent probes appear as diffraction-limited dots within individual cells. Each dot, produced by collective binding of probes to target transcript, indicates a single RNA molecule (Femino et al., 1998, Raj et al., 2008). In analyses of >20,000 WT cells, FLO11 dots were detected in 69% of cells (±1.6 standard error of the mean (SEM) calculated from four experiment replicates), while the remaining 31% (±1.4 SEM) of cells were devoid of FLO11 dots (Figures 1D and and3D;3D; Table 1).
The FISH microscopy images provide quantitative information about alternative FLO11 expression states. We observe subpopulations of WT cells that exhibit no FLO11 transcripts (0 dots), low-copy basal-level transcription (1 to 5 dots), or high-copy active transcription (>5 dots per cell; Figures 1D and and3D).3D). Of cells in which FLO11 transcripts are detected, 30% (±0.7 SEM) exhibit basal-level transcription and 39% (±0.8 SEM) are active for FLO11 transcription, reproducibly containing ~30 mean transcripts per cell (Table 1).
Alternative FLO11 expression states are also present in null mutants for flo8, Rpd3L− (i.e., cti6), and sfl1. Most flo8 and cti6 cells either contain no FLO11 transcripts or exhibit basal-level transcription independent of these trans-activators (Figure 1D). Basal transcription is insufficient to support Flo11-dependent colony morphology (Figures 3C and and4D),4D), adhesion, or filamentation (Bumgarner et al., 2009). In sfl1, 98% of cells are active for FLO11 transcription (Figure 1D), containing on average 36 transcripts (± 0.2 SEM) per cell (Table 1). The active cells in WT and sfl1 populations contain similar numbers of FLO11 transcripts (Table 1), suggesting that the overexpression of FLO11 noted in population-based studies of sfl1 (Pan and Heitman, 2002; Halme et al., 2004) is mainly due to an increase in the number of active sfl1 cells rather than an increase in the number of transcripts per cell.
PWR1 and ICR1 were also imaged in WT, flo8, cti6, and sfl1 cells (Figures 1C–1D and S1). The ncRNAs are observed within nuclei and in cytoplasm (Figure 1C and S1). They are detected in some cells within each clonal population but are completely absent in others, revealing variegation of PWR1 and ICR1 (Figure 1D). When either PWR1 or ICR1 is observed, we detect on average fewer than two transcripts per cell (Table 1), which may indicate intermittent transcription, low transcription rates, short half-lives, and/or technical limitations in our method of detection. Data collected from FISH experiments are largely consistent with previous observations from northern blots (Bumgarner et al., 2009). PWR1 is detected more frequently in sfl1 cells than in WT (Figure 1D), whereas ICR1 is detected less often in sfl1 cells than in WT (Figure 1D). In cti6 and flo8, the percentage of cells containing PWR1 is lower relative to both WT and sfl1 (Figure 1D). ICR1 is detected more often in cti6 and flo8 cells than in sfl1, but only the cti6 mutant shows an increase in the percentage of cells in which ICR1 is detected relative to WT (Figure 1D).
FLO11 and ncRNAs were imaged simultaneously using spectrally distinct fluorophores. The coincidence of FLO11 and PWR1 transcripts in individual cells supports PWR1’s role in promoting FLO11 expression (Figure 1C and S1A). In active WT cells (i.e., cells containing >5 FLO11 dots), there is a strong positive correlation between PWR1 and FLO11. As PWR1 count increases, the mean and median FLO11 count also increases (Figure 2A). This observation deviates significantly from results expected under a null hypothesis in which PWR1 and FLO11 counts are independent (i.e., where PWR1 count predicts no change in FLO11 count, β = 0). Instead, using a linear regression model, the results are consistent with each PWR1 dot predicting 8 additional FLO11 transcripts in a given cell (β = +8.2 FLO11 per PWR1, 95% confidence interval (CI) = +6.9 to +9.5, p-value = 1.99E-34).
Conditional on detection of PWR1 in a given cell, the probability that the cell is also active for FLO11 transcription is significantly higher than predicted under the null hypothesis (Table 2). To underscore this relationship, since ~40% of WT cells are active for FLO11 transcription (Figures 1D and and3D),3D), we would expect under the null hypothesis to find that ~40% of PWR1-positive cells are also active for FLO11. Instead, ~90% of PWR1-positive cells detected are also active for FLO11 transcription, supporting a positive correlation (p-value = 8.71E-65) between PWR1 transcription and FLO11 transcription (Table 2).
Simultaneous imaging of FLO11 and ICR1 in single cells supports ICR1’s role in repressing FLO11 expression (Figures 1C and S1B). Mean and median FLO11 dot counts decrease as ICR1 dot count increases (Figure 2B). Linear regression analysis of the full data set (Figure 2B) yields a model in which each ICR1 dot predicts 2 fewer FLO11 dots in a cell (β = −2.1 FLO11 per ICR1, 95%CI = −1.0 to −3.1, p-value = 1.54E-04). The presence of 2 or more ICR1 dots is coincident with marked reduction in FLO11 (Figure 2B). Linear regression analysis of the subset of cells in which 2 or more ICR1 dots are detected (Figure 2B) predicts 4 fewer FLO11 dots per ICR1 (β = −4.1 FLO11 per ICR1, 95%CI = −1.3 to −6.9, p-value = 0.0049). FLO11 and ICR1 dots are coincident in some cells (Figures 1C and S1B), but it is not possible to discern from our data whether these cells were actively transcribing both transcripts or had undergone a recent switching event.
Simultaneous imaging of PWR1 and ICR1 ncRNAs in individual cells supports the existence of a ncRNA toggle (Bumgarner et al., 2009; Figures 1C, ,2C,2C, and S1C). Under a null hypothesis in which these ncRNAs are independent, the number of PWR1 dots is not expected to correlate with the number of ICR1 dots. Instead, we observe a significant decrease in the mean number of ICR1 dots detected as the number of PWR1 dots increases in WT and sfl1 cells (Figure 2C). This effect is observed when cells are binned according to PWR1 count and then mean and 95%CI are determined for each binned population’s ICR1 counts (Figure 2C). Linear regression performed on the full set of WT cells (Figure 2C) shows that each PWR1 dot predicts 1 fewer ICR1 transcript within a given cell (β = −0.9 ICR1 per PWR1, 95%CI = −0.7 to − 1.1, p-value = 5.86E-22). Analysis of the subset of WT cells that contain either 0 or 1 PWR1 dot (i.e., comparing the bins between which the greatest change in ICR1 is observed (Figure 2C)) reveals a marked reduction in ICR1 count predicted by the presence of PWR1 (β = −1.2 ICR1 per PWR1, 95%CI = −1.0 to − 1.5, p-value = 1.84E-21). Linear regression performed on the full sfl1 population summarized in Figure 2C shows that each PWR1 dot predicts 1 fewer ICR1 transcript within a cell (β = −0.8 ICR1 per PWR1, 95%CI = −0.7 to −1.0, p-value = 1.30E-25). When only the subset of sfl1 cells that contain either 0 or 1 PWR1 dot are analyzed, an even greater reduction in ICR1 is predicted by the presence of PWR1 (β = −1.9 ICR1 per PWR1, 95%CI = −1.6 to −2.2, p-value = 1.09E-39). ICR1 and PWR1 are sometimes observed together in cells (Figures 2C and S1C), which may be indicative of recent switches of the toggle (i.e., where one ncRNA is being newly synthesized while the other persists because it has not yet been degraded).
Transcription of ICR1 was reduced via three distinct methods in the cti6 mutant. One method uses a transcriptional terminator to disrupt ICR1 (icr1::Term, T3 in Bumgarner et al., 2009). Another (cti6 ΔpICR1) reduces ICR1 transcription by removing the ncRNA’s upstream regulatory sequences (Figure S2). The third (cti6 pMET-ICR1) controls ICR1 under the MET25 promoter (Figures 3 and and4),4), which is repressed in rich media and induced in media lacking methionine.
Decreasing ICR1 transcription by any of these approaches increases FLO11 expression in bulk-cell assays (Figure 3A) and restores Flo11-dependent colony morphology (Figure 3C). In all three cases, FISH experiments show that reduction of ICR1 recovers cells active for FLO11 transcription (Figures 3C and 3D). The number of FLO11 transcripts detected in rescued active cells (~30 dots per cell) is similar to the number observed in active WT and sfl1 cells (Table 1). Thus, reduction of ICR1 transcription in the cti6 mutant restores a subpopulation of active cells that is indistinguishable in quality, although different in population frequency, compared to active cells observed in WT.
Bulk-cell assays reveal that average FLO11 levels are elevated but not fully returned to wild type when ICR1 transcription is reduced in the cti6 background (Figure 3A). Several models could explain this observation (Figure 3B). In Model 1, the same percentage of cells is “on” in WT and the rescued strain, but WT cells express FLO11 more highly. In Model 2, a smaller percentage of cells turns “on” in the rescued population, but each rescued cell expresses FLO11 at a level similar to WT active cells. In Model 3, reduction of ICR1 enables all cells in the rescued population to express FLO11, but each at a very low level. Single-cell imaging enabled distinction among these models, showing that Model 2 is most appropriate to explain the observed phenomena. These results suggest that an additional ICR1-independent repressor is also dysregulated in the cti6 mutant. The additional repressor may be Sfl1, which shows enriched recruitment to the FLO11 promoter in the cti6 mutant compared to WT (Figure 5D).
When ICR1 transcription is reduced in cti6, PWR1 is detected in a higher percentage of cells compared to the unmodified cti6 cell background (Table 2). Conditional on detection of PWR1 in a given cell, the probability of high-copy FLO11 transcripts being detected in that same cell is significantly higher than expected if PWR1 and FLO11 transcription were independent events (Table 2). For example, 14% of cti6 ΔpICR1 cells are active for FLO11 transcription (Table 2). Thus under the null hypothesis, 14% of PWR1-positive cells would be expected to be active for FLO11 transcription. Instead, we see that 68% of PWR1-positive cells are active for FLO11 transcription (mean dot count: 35 ± 0.6 SEM; Table 1), a significant 4.7-fold increase over the expectation under the null (p-value = 1.80E-97) (Table 2).
An examination of the shifting distributions of alternative FLO11 expression states (Figure 3D) suggests that, when ICR1 transcription is reduced in cti6, the rescued subpopulation of active cells may be derived mainly from the subpopulation of basal cells. This observation raises the possibility that, rather than playing a direct role in modulating silencing of FLO11 transcription, the ncRNA toggle plays a role in the switch from basal to active state. This idea is further supported by northern analysis that shows that the Hda1 HDAC does not affect ICR1 transcription (Figure S4F), suggesting that Hda1-mediated silencing (Halme et al., 2004; Octavio et al., 2009) occurs downstream or independently of ICR1-mediated FLO11 repression.
To investigate further the effect of ICR1 on FLO11 expression, heterologous promoters (Janke et al., 2004) were inserted to control ICR1 transcription. Increased ICR1 transcription under TEF (pTEF) or GPD1 (pGPD) promoters results in decreased FLO11 in WT and sfl1 (Figures 4A–4C) and loss of Flo11-dependent colony morphologies—a particularly striking result for the sfl1 mutant which normally produces very crinkly colonies (Figure 4D). Conversely, reduction of ICR1 under the MET25 promoter (pMET), which is repressed in YPD, results in elevated FLO11 transcript levels and restores crinkled colony morphology to the cti6 mutant (Figure 4). In contrast, when pMET-ICR1 strains are grown in synthetic media that lacks methionine (i.e., when the pMET promoter is induced), we observe the inverse effect: a reduction of FLO11 transcript levels (Figure 4C).
The distribution of FLO11 counts detected in the WT population can be reconstituted by combining the distributions observed in flo8 and sfl1 mutants (Figure 5A). Furthermore, recruitment of Flo8 to the FLO11 promoter is reduced in the cti6 mutant and increased in the sfl1 mutant (Bumgarner et al., 2009). These results provoked further examination of the relationship between ICR1 transcription and the recruitment of key trans-acting factors to the FLO11 promoter.
ICR1 transcription inhibits recruitment of Flo8 and Sfl1 to the FLO11 promoter. When ICR1 transcription is reduced in the cti6 mutant (Figure 3), there is a marked increase in the recruitment of Flo8 (Figure 5B) to its binding region (Pan and Heitman, 2002). ChIP performed on strains in which heterologous promoters increase ICR1 transcription result in reduced recruitment of myc-tagged Flo8 and Sfl1 to the FLO11 promoter (Figure 5C–5D). Repression of ICR1 transcription under pMET when WT and mutant strains are grown in rich media results in the enrichment of myc-tagged Flo8 and Sfl1 to the FLO11 promoter (Figures 5B–D). These data demonstrate that ICR1 transcription interferes with recruitment, occluding (Martens et al., 2004) or ejecting these trans-acting factors from the FLO11 promoter.
We developed a computational framework that captures the changes in measured FLO11 transcript distributions observed across genotypes as a function of recruited Flo8 transcription factor. A mixture model (McLachlan and Peel, 2000) assumes two populations of cells, one with no Flo8 recruitment that exhibits basal/low FLO11 expression and another with maximum Flo8 recruitment that exhibits high-copy active FLO11 expression. The parameters for the population of cells with no Flo8 recruitment were determined empirically using the flo8 deletion strain. The FLO11 transcript distribution in flo8 is best fit with a Poisson distribution using maximum likelihood optimization (Figure 5E). The parameters for the population of cells with maximum Flo8 recruitment were determined using the sfl1 deletion strain. In sfl1, a Gamma distribution is the best fit for the measured FLO11 transcript distribution (Figure 5E). Once parameters were determined from these fits, the mixture model was constrained to one free parameter, namely the fraction of cells exhibiting high-copy FLO11 transcript expression. We fit the mixture model to FLO11 transcript distributions observed in WT, cti6, cti6 ΔpICR1, and cti6 pMET-ICR1 strains (Figure 5E and S3). A strong positive correlation (Figure 5F) is observed between the amount of Flo8 recruitment measured by ChIP and the fraction of cells exhibiting high-copy FLO11 transcripts within a given population. This combination of experimental and computational approaches supports the hypothesis that the ncRNA ICR1 modulates alternative FLO11 expression states by controlling Flo8 recruitment to the FLO11 promoter (Figure 5G).
Single-cell resolution FISH imaging has revealed alternative FLO11 expression states that were not detectable by other methods (Halme et al., 2004; Bumgarner et al., 2009; Octavio et al., 2009) and directly demonstrate that FLO11 mRNA itself variegates in clonal populations (Figures 2–3 and S1). In WT, one class of cells is devoid of FLO11 transcripts, suggesting transcriptional inactivity or silencing at FLO11. A second class contains 1 to 5 transcripts per cell, exhibiting low-level basal FLO11 transcription. The third class is active for FLO11 transcription, with a mean count of ~30 FLO11 transcripts per cell (Table 1). Thus, Flo11-dependent phenotypic heterogeneity observed in WT clones results from substantial cell-to-cell differences in FLO11 expression rather than noisy low-level expression across the population.
The three classes of cells may represent alternative promoter states predicted computationally to exist at FLO11 (Octavio et al., 2009) and demonstrated experimentally at other loci (Vermaak et al., 2003; Li et al., 2007): a silent promoter state mediated by local chromatin structure, a competent but inactive or basal promoter state resulting from absence of required trans-activators or presence of trans-acting repressors, and an active promoter state. The importance of such alternative states in cellular differentiation is clear for multicellular organisms, composed of genetically homogeneous cells that are structurally and functionally heterogeneous due to differential gene expression. These alternative expression states also have biological significance for clones of unicellular yeast. They explain the phenomenon of Flo11-dependent phenotypic variegation (Halme et al., 2004) that may provide a survival advantage, not to individual cells per se, but to the clone’s genetic identity by promoting survival in fluctuating environmental conditions (Buttner et al., 2006; Batada and Hurst, 2007; Acar et al., 2008; Lehner, 2008).
The distinction discerned between the basal (1 to 5 dots) and active (>5 dots) expression states of FLO11 is biologically meaningful. Flo8 is recognized as the key activator for FLO11, and null alleles of flo8 or cti6 exhibit loss of Flo11-dependent phenotypes such as haploid adhesion, crinkly colony morphology, and diploid filamentation (Figures 3 and and4;4; Liu et al., 1996; Guo et al., 2000; Bumgarner et al., 2009). Yet flo8 and cti6 mutant populations contain many cells that exhibit basal-level expression (Figure 1D), demonstrating that ≤5 copies of FLO11 per cell is insufficient to support Flo11-dependent phenotypes.
Single-cell resolution has revealed mechanistic aspects of the regulatory circuitry at FLO11 that would not have been discernable using population-wide measurements. Reduction of ICR1 transcription in the cti6 mutant causes a subset of cells to recover active transcription (Figure 4C and 4D), pointing to a causal role for ICR1 in repressing active FLO11 expression in individual cells. Together, empirical results and computational modeling suggest that ICR1’s repressive effect is due to occlusion or ejection of key trans-acting factors Flo8 and Sfl1 ((Figure 5; Martens et al., 2004; Bumgarner et al., 2009) from their respective binding sites on the FLO11 promoter (Pan and Heitman, 2002). Since PWR1 is not detected in every cell that is active for FLO11 transcription (Figure 2A) and ICR1 is not detected in every cell that is “off” for FLO11 (Figure 2B), these ncRNAs might not be required to maintain alternative FLO11 transcription states but could instead help transition the locus between states.
Previous results (Bumgarner et al., 2009) show that ICR1 and PWR1 exert their effects on FLO11 and on each other via a cis-acting process. Thus, the process of transcription rather than the products of the transcriptional process are mechanistically important for the toggle. Transcription of ICR1 along the length of the FLO11 promoter may serve to “reset” the promoter by transiently eliminating interactions between the DNA and trans-acting activators and repressors, such that Flo8 and Sfl1 compete anew for recruitment to the FLO11 promoter (Figure 5G). ICR1 transcription may thereby influence the likelihood of downstream events that lead to an active or inactive FLO11 transcription state. The competitive binding of Sfl1 or Flo8 (Pan and Heitman, 2002) is also central to the toggle. Their recruitment is influenced by the activity of Rpd3L (Figure 5) and feeds back to determine which ncRNA transcript program is initiated. Sfl1 initiates a cascade of events that result in reversible transition to the silenced state (Halme et al., 2004), whereas Flo8 initiates events that transition the FLO11 promoter from basal to active state. Our studies suggest that recruitment of Flo8 induces a pulse of PWR1 transcription that promotes the FLO11 active state by interfering in cis with ICR1 transcription (Figure 5G).
Quantitative RNA FISH assays in single cells, genetic analysis, and computational modeling together have power to provide unanticipated insights into the cis-acting roles of ncRNAs. The integration of experimental techniques used in this study has enabled a quantitative understanding of the function of long ncRNAs in gene regulation in yeast and may prove to be a useful strategy for investigating these transcripts across organisms.
Yeast strains (See table provided in Supplemental Experimental Procedures) were derived from Σ1278b (Liu et al., 1996). Standard media were prepared and genetic manipulation techniques were carried out as described (Guthrie and Fink, 2001). Deletions of the endogenous ICR1 promoter region were generated as described in Figure S2 (Güldener et al., 1996). NatR-marked promoters pTEF (pYM-N19), pGPD1 (pYM-N15), and pMET25 (pYM-N35) were integrated 3446 bp upstream of the FLO11 ATG, without loss of endogenous sequence, to control ICR1 (Janke et al., 2004). For northern blot analysis, qPCR, and ChIP, cells were grown overnight at 30°C in YPD liquid, diluted to OD6000.1, and all cultures grown to either OD6000.8–1.2 or OD6002.8–3.0.
RNA FISH was performed as described (Raj et al., 2008) with the following modifications: Yeast cultures were grown at 30°C in YPD liquid from starting concentration OD6000.1 to final concentration OD6002.8–3.0. Formaldehyde fixation was performed for 30 minutes at 22°C and continued overnight at 4°C, with gentle rocking throughout. Zymolyase digestions were performed at 30°C in TV 500ul Buffer B containing 8ul zymolyase (2.5mg/ml) for 1.25 hours while rotating tubes. Hybridizations with DNA probes (Figure 1A and table provided in Supplemental Experimental Procedures) were performed in 10% formamide hybridization buffer. FLO11-specific probes were coupled to TMR, PWR1-specific probes were coupled to Cy5, and ICR1-specific probes were coupled either to TMR or Cy5. To protect fluorophores from oxidation during imaging, cells were suspended in GLOX buffer as described (Raj et al., 2008) and imaged on standard glass microscope slides using cover slips sealed with silicon gaskets.
Images were collected using a Nikon TE2000 inverted fluorescence microscope with 100x oil-immersion objective, appropriate filters (TMR, Cy5, and DAPI), and a Princeton Instruments camera with MetaMorph software (Molecular Devices, Downington, PA). Custom filter sets were designed to distinguish TMR and Cy5 signal. Differential Interference Contrast (DIC), DAPI, TMR, and Cy5 images were collected with 0.2 micron Z-slices. DIC and DAPI images were used to identify individual cells. TMR and Cy5 image stacks were used to detect RNA transcripts. For image processing, a DIC image was chosen in which a clear cell boundary could be observed. This image was converted into a binary image using automated thresholding. The maximum projection of a DAPI-image stack was generated and converted into a binary image using a fixed pixel intensity threshold. The binary DIC image was merged with the binary DAPI image. DAPI-stained nuclei were used in running a marker-controlled watershed algorithm over the merged DIC/DAPI image. Cell boundaries of individual cells were obtained using an edge detection algorithm. Connected regions measuring larger than the expected range of sizes for an individual cell were rejected. The number of RNA transcripts in each cell was counted using a program that operates as follows: to enhance particulate signals, the program runs a median filter followed by a Laplacian filter on each optical slice. A threshold was then selected to detect individual dots in each plane. The particle count was robust over a range of selected thresholds. Images that demarcated cell boundaries were merged with each plane of TMR or Cy5 image stacks. This processing enabled the program to count the total number of isolated signals in three dimensions within each cell.
Total RNA was isolated by standard acid phenol extraction and oligo(dT) selected (Qiagen Oligotex mRNA Kit) to enrich for polyadenylated transcripts. RNAs were separated on formaldehyde-agarose denaturing gels and blotted as described (Sambrook et al., 1998). Hybond membranes (Amersham) were hybridized with strand-specific 32P-labeled RNA probes generated using the Ambion T7 Maxiscript Kit. For load controls, a 32P (exo-) Klenow-labeled DNA probe specific to transcript SCR1 was used, with the exception of the blot in Fig. S4F, in which a 32P (exo-) Klenow-labeled DNA probe specific to transcript TPI1 was used.
Total RNA obtained by standard acid phenol extraction was reversed transcribed (Qiagen QuantiTect Kit). cDNAs were analyzed with specific primers, SYBR Green reagents (Applied Biosystems), and the ABI 7500 qPCR system.
Protocols have been described (Lee et al., 2006). Briefly, IPs were performed with Dynal Protein G magnetic beads preincubated with antibody against Myc-epitope (Covance 9E-11 MMS-164P). SYBR Green qPCR (Applied Biosystems) was performed on IP and WCE with specific primers.
For regression analyses, where FLO11 transcript count (the outcome variable) was regressed against PWR1 or ICR1 transcript number (the predictor variable), a log-additive model relating the predictor to the outcome was assumed. Linear regression was performed with the statistical software package R using the glm() function. For the other tests of independence between the transcripts, a standard Pearson’s chi-square test was performed.
Our approach assumed a simple mixture model of a Poisson and a Gamma distribution (McLachlan and Peel, 2000). The Poisson distribution consists of one parameter, the normalized basal transcription rate λ = 0.64 mRNA. The Gamma distribution consists of two parameters, the mean number of mRNA transcripts produced at each burst (i.e., average burst size; Raj et al., 2006) θ = 9.5 and the normalized deactivation rate k = 3.9. After the rates for these two distributions were determined, we fit the remaining FLO11 mRNA distributions with the mixture model,
where F is the fraction of cells within a given population that exhibit high-copy (active) FLO11 mRNA expression. The fit of the mixture model to the observed data was assessed using a maximum likelihood approach.
We thank Professors Rick Young (Whitehead Institute) and Joe Heitman (Duke University) for reagents to generate myc-tagged Sfl1; Chia Wu for strains yCW91 and yCW180; Leah Octavio for insightful comments on the manuscript; and Garrett Hauck, Lisa Nguyen, and Rafael Widjajahakim for help with media preparation and DNA isolation. This work was supported by National Institutes of Health Grants GM035010 (G.R.F), GM40266 (G.R.F), and 1DP1OD003936 (A.v.O.), by National Science Foundation Grant ECCS-0835623 (A.v.O.), and by the Deutsche Forschungs Gemeinschaft Forschungs Stipendium (G.N.). G.R.F. is an American Cancer Society Professor.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.