Final report on work performed and results achieved
January 2006 - December 2009
1. Methodologies
In order for data from different laboratories to be compared in the models, it is important that identical techniques and reagents are used. A set of standard operating procedures was achieved by discussion, sharing written protocols and by the exchange of personnel and reagents. A Wiki server was set up to allow all the partners to access and contribute to the RiboSys project documentation.
Initial work by the biologists focused on refining and optimising the following protocols for use with yeast:
- Rapid sampling (at 10 second intervals) of yeast cell cultures for RNA extraction
- Optimum extraction of both pre-mRNA and pre-rRNA from yeast cells
- Calculation of cell number based on optical density and direct cell counting methods
- Determining the efficiency of extraction of RNA and quantifying transcripts per cell
- Northern blotting
- Metabolic labelling of pre-rRNA
- Metabolic labelling of pre-mRNA
- Cap-specific cloning of de novo RNAPII transcripts
- Primer extension analysis
- Chromatin immunoprecipitation (ChIP)
- Quantification by reverse transcription-quantitative real-time PCR (RT-QPCR)
- Splicing by RT-QPCR
- Transcription run-on
- RNase H analysis for high-resolution northern gels
- Poly(A) Status Estimation (PASE)
- Visualisation of single transcripts in yeast cells with fluorescence imaging techniques.
Detailed protocols are available on the protocol page.
2. Pre-messenger RNA transcription and processing studies
2.1. Yeast strains
Three yeast strains were produced for tetracycline/doxycyclin repressible (tetOFF) expression of reporter genes, based on CEN.PK2, BY4747 and W303 strains with all components integrated in the genome. The strains based on W303 were selected for further work. There are seven variants of the reporter genes, which are hybrids of ACT1/PGK1: Ribo1 contains an efficiently spliced derivative of the ACT1 intron; ILRibo1 is intronless, 5′SSRibo1, BSRibo1 and 3′SSRibo1 are splicing defective variants containing mutations at the 5′splice site, branch site or 3′ splice site respectively, and Ribo500 and Ribo200 have the second exon extended by 500 and 2000 nucleotides respectively.
Tetracycline/doxycyclin inducible (tetON) yeast strains constructed using the same approach of integrating all components in the genome did not work and a new approach had to be pursued. The best solution found was to produce the tTA′ transactivator protein from a strong promoter on a plasmid (Alexander, Barrass, Dichtl ,et al., manuscript submitted).
2.2. RNA copy number measurements
Fluorescent in situ hybridisation (FISH) was used successfully to detect individual reporter transcripts in single yeast cells using specially designed probes. Using FISH we have measured the number of individual mRNA molecules per cell in large cell populations for the Ribo1 reporter and several mutant derivatives. These distributions were fitted to various models, and the results suggest that transcription initiation is more complex than a single step stochastic model, and undergoes ON/OFF transitions.
Average copy numbers per cell of the Ribo1 reporter transcripts in the induced and uninduced tetON and tetOFF strains were estimated using RT-QPCR, and the results were validated by comparison with the FISH direct visualisation approach. Variation in measurements was found to be largely due to biological variation between different cultures and is likely explained by the huge cell to cell variation. (Alexander, Barrass, Dichtl, et al., manuscript submitted).
2.3. Transcription, splicing and decay
Using rapid sampling techniques and RT-QPCR the kinetics of transcription and splicing were measured during reporter gene induction. RNA polymerase (RNAPII) and splicing factor recruitment to the gene were assayed by chromatin immunoprecipitation (ChIP). TetOFF strains were used to measure RNA decay rates in wild-type cells and in mutants defective in RNA degradation machinery, and RNA synthesis rates were estimated.
Modelling and simulations of these kinetic datasets produced models with good fit to the data. Model predictions were made to explain the kinetics of pre-mRNA and spliced mRNA accumulation. These were tested experimentally using RNA degradation mutants, longer reporter genes and by measuring co-transcriptional splicing.
An important outcome of this work was the observation that splicing of the reporter transcripts takes place largely co-transcriptionally (i.e. at the DNA template prior to completion of the transcript). A previously undetected influence of splicing on the kinetics of transcription was observed. This suggests that transcription and splicing are functionally coupled in yeast, possibly through a splicing-dependent transcriptional checkpoint (Alexander, Innocente, et al., submitted).
2.4. mRNA 3′ end formation
Ribo1 reporter transcription and polyadenylation were measured following interference with small molecule inhibitors (mycophenolic acid and cordycepin). The presence of either drug strongly reduced transcription, with the onset of transcription being significantly delayed. Cordycepin also resulted in increased accumulation of unspliced pre-mRNA indicating that it may interfere with pre-mRNA splicing.
The analysis of GAL7 expression and polyadenylation was done in wild-type strains and in isogenic mutants defective in various steps of the gene expression pathways. Deadenylation kinetics were measured and steady-state polyadenylation determined. While the absence of the major deadenylase strongly affected deadenylation rate, we observed only minor effects with mutants affecting transcriptional elongation and chromatin structure.
The major insights emanating from this work are: 1) Turn-over rate of the reporter transcripts was not determined by the rate of poly(A) tail removal. 2) The history of the reporter mRNA (e.g. is it spliced or not) does affect the fate of the mRNA as seen with the rate of mRNA deadenylation. 3) During transcriptional activation mRNAs are licensed for stable accumulation and possibly engagement with the translational machinery. The mechanism is, however, currently unknown.
2.5. Modelling - mRNA
The required methods to construct and simulate kinetic models of mRNA processing were selected and developed. Two different approaches were investigated: (i) solving ordinary differential equations (ODE) systems both numerically using a propriety software package (based on Matlab's numerical analysis and optimization tools), and analytically whenever possible (i.e., in the linear case), and (ii) stochastic simulations using the Gillespie algorithm as implemented in the Dizzy tool. In order to obtain experimentally-verifiable models, the software was developed to analyse experimental data, to compare it with the model predictions, and to assess model parameters by optimising the agreement between the two. The software provides estimates of model parameters and confidence levels of these values (both by themselves and jointly).
The approaches mentioned above were tested on small model systems, such as transcript elongation, and then applied for describing the kinetics of transcription and splicing of the Ribo1 reporter gene and several of its variants in the tetOFF and tetON strains. Several qualitative and quantitative predictions were made based on these models regarding the underlying mechanisms that can produce the observed behaviour. These models included transcription initiation, elongation and splicing and were fitted to the experimental measurements and refined in an iterative manner as more data became available. In particular, these models allowed the first step of splicing to be either co- or post-transcriptional, i.e. occurring before or after transcription termination, and hence to elucidate the kinetics of the splicing process.
The distribution of the Ribo1 mRNA copy number in individual cells was also modelled using both the ODE and stochastic approaches. The former approach provided insight regarding the type of reactions that can produce the observed distribution. These include several variants of a two-stage creation-degradation or activation-deactivation model with a hidden first stage. One of these variants is equivalent to the random telegraph model, and the hidden first stage can be interpreted as the promoter undergoing On/Off transitions. The stochastic model was explored by Monte-Carlo simulations using the Gillespie algorithm, and the results were fitted to the experimental data in order to estimate the rates of On-Off transitions and transcription initiation. The resulting fit utilised high moments of the distribution, up to and including the fourth one. Variants of the random telegraph model were constructed as well, representing the elongation process in greater detail, and exploring the possibility that RNAPII complexes are processive, i.e. continue to synthesise mRNA even if the gene switches to the Off state, and, conversely, that RNAPII is coupled to the promoter state (Aitken et al., PLoS ONE (2010) 5:e8845).
3. Pre-ribosomal RNA transcription and processing studies
3.1. Pre-rRNA transcription
Values were obtained for rRNA transcription in wild-type and mutant strains using a standardised transcription run-on approach. These permitted evaluation of the impact of defects in rRNA synthesis/processing pathway, such as transcription elongation and termination, processing or silencing, on the overall rates of transcription. These experiments demonstrated that transcription elongation factor Spt4 that is common to RNA polymerases I (RNAPI) and II (RNAPII) contributes to the transcription rate of both polymerases, but more so for RNAPII than for PolI. The effect of the small subunit of RNAPI, Rpa12, that participates in RNAPI transcription termination, is the opposite; cells lacking this component synthesise rRNA considerably faster. This effect suggests that Rpa12 may function in the regulation of RNAPI velocity, possibly to facilitate proofreading. Datasets of rRNA transcription rates in mutants affecting transcription point to an important contribution of polymerase processivity to rRNA synthesis. These two aspects, processivity and proofreading, are additional factors contributing to the overall performance of the polymerase. Analogous analyses performed in different growth conditions showed surprisingly that rRNA transcription efficiency does not directly correlate with cell growth rate, but may correlate with the optimal growth for cell survival. Through a metabolic labelling approach, the transcription time of the 35S primary transcript in vivo was determined to be ~3min. This is a key parameter for all modelling of the data. Previous estimates were based on in vitro transcription rates or indirect estimates from "Miller" chromatin spreads.
During transcription termination by RNAPII on protein-coding genes, the nuclear 5′ exonuclease Rat1 degrades the nascent transcript downstream from the polyadenylation site and "torpedoes" the transcribing polymerase, leading to termination. We found that the activity of Rat1 is also required for efficient termination by RNAPI on the rDNA. In strains lacking catalytically active Rat1 or its cofactor Rai1, RNAPI reads through the major, "Reb1-dependent" terminator (T1) but stops downstream at the "fail-safe" terminator (T2) and replication fork barrier (RFB). We propose that co-transcriptional cleavage of pre-rRNA by the Rnt1 endonuclease produces a loading site for the Rat1/Rai1 complex, which then degrades the nascent transcript. When Rat1 catches RNAPI, transcription is terminated (El Hage et al., 2008, Genes Dev. 22:1069-81).
3.2. Pre-rRNA processing
To generate kinetic data for pre-rRNA processing at an appropriate time scale, we have developed new techniques for the rapid harvesting (down to 10 sec) of metabolically labelled yeast cultures. Incorporation of tritium label into pre-rRNA and rRNA species was quantified by gel electrophoresis and imaging.
3.3. Modelling - rRNA
Modelling of the kinetic data allowed comparison of predicted pre-rRNA lifetimes and processing patterns with the experimental data. This supported a transcription time for the 35S primary transcripts of ~170 sec at 30°C, with a high level (~70%) of cleavage of the nascent transcripts at the early processing sites that generate the 20S precursor to the 18S rRNA. Yeast pre-rRNA modification was previously reported to take place exclusively on released transcripts but the high degree of nascent transcript cleavage predicted by the model strongly indicated that modification also occurred on nascent transcripts. Applying a second round of high resolution kinetic labelling, we showed that the pre-rRNA does indeed undergo substantial nascent transcript methylation. (Kos and Tollervey, 2010, Mol. Cell, in press). In alternative approach, a "geometrical" model was used (Plyusnina, Speshilov, Strizh, Kos, Tollervey, Goryanin and Demin, submitted). This predicted the hypothetical existence of additional long-lived pre-rRNA, which has not yet been identified in the experimental data.
We have successfully applied mathematical modelling analyses to understanding the pre-rRNA processing pathways. Our demonstration that pre-rRNA processing and modification both largely occur on the nascent transcripts represents a substantial change from the long-established conventional view in the field that these are exclusively post-transcriptional activities.
4. Conclusion
Despite the large numbers of previous molecular genetic analyses, many outstanding questions remain concerning the pathways of messenger RNA and ribosomal RNA transcription and processing in budding yeast. The findings from this project are a clear demonstration that high resolution, quantitative analyses combined with mathematical modelling can provide biological insights that were not obtained by more "conventional" approaches. In addition, the project provided new yeast strains, protocols and novel insights that will facilitate future research in this area.