|Home | About | Journals | Submit | Contact Us | Français|
Messenger RNA decay measurements are typically performed on a population of cells. However, this approach cannot reveal sufficient complexity to provide information on mechanisms that may regulate mRNA degradation, possibly on short time scales. To address this deficiency, we measured cell cycle regulated decay in single yeast cells using single-molecule FISH. We found that two genes responsible for mitotic progression, SWI5 and CLB2 exhibit a mitosis-dependent mRNA stability switch. Their transcripts are stable until mitosis when a precipitous decay eliminates the mRNA complement, preventing carry-over into the next cycle. Remarkably, the specificity and timing of decay is entirely regulated by their promoter, independent of specific cis mRNA sequences. The mitotic exit network protein, Dbf2p binds to SWI5 and CLB2 mRNAs co-transcriptionally and regulates their decay. This work reveals the promoter-dependent control of mRNA stability, a novel regulatory mechanism that could be employed by a variety of mRNAs and organisms.
Precise analysis of decay kinetics is necessary to understand when and how a decay regulator functions and single-cell, single-molecule techniques could advance our understanding of mRNA turnover. For example, the kinetic behavior of individual RNA polymerase II (RNAPII) transcribing a gene (reviewed in (Ardehali and Lis, 2009)) provides a precise quantification of the contribution of mRNA synthesis to the cellular pool of transcripts. However, to date no such approach has been available for measuring mRNA turnover. Traditional techniques have relied on normalization of decay signal and on a large sample of cells, genetically modified or treated with inhibitors, to stop transcription and thus obtain kinetic information of a decaying mRNA species (reviewed in (Passos and Parker, 2008)). Furthermore, the accuracy of decay measurement varies with the technique used. For example, in budding yeast, half-lives of an individual mRNA species quantified by different approaches may differ by more than 50 percent (Grigull et al., 2004; Holstege et al., 1998; Wang et al., 2002). In turn, the accuracy of the decay curve will influence how precisely it can be modeled. In this work we use single-molecule counting with Fluorescent In Situ Hybridization (FISH) (Zenklusen et al., 2008) to derive an absolute measure of mRNA synthesis and decay in individual cells. This provided a highly sensitive approach for detecting changes occurring in a fraction of cells, that otherwise would have been obscured.
We focused on mRNA turnover since it could regulate gene expression during the cell cycle. For instance, entry into mitosis induces a rapid mRNA decay of the mitotic Clb2p cyclin that, if prevented, can cause failure of cells to finish mitosis (Cai et al., 2002; Gill et al., 2004). Entry into G0 causes stabilization of specific G0 mRNAs (Talarek et al., 2010), whereas the stability of the canonical histone mRNAs increases with the onset of S phase and exit from S phase induces their rapid decay (Marzluff et al., 2008; Osley, 1991). Thus, together with their cyclical transcription, the destabilization of mRNAs can restrict the activity of periodically expressed genes to a particular cell cycle phase. This modulation of stability is typically achieved through binding of decay regulators to specific sequences located in the mRNA (reviewed in (Guhaniyogi and Brewer, 2001)).
We focused on two cell cycle regulated genes, SWI5 and CLB2 and measured changes in their mRNA turnover during the cell cycle. Swi5p is a transcription regulator of late mitosis genes and Clb2p is a G2 phase cyclin that drives the progression of cells towards mitosis. They are co-regulated through shared promoter elements (Koranda et al., 2000; Spellman et al., 1998; Zhu et al., 2000) and were measured to degrade with 8 min and 4.5 min half-lives, respectively (Wang et al., 2002). We used morphological markers to determine timing of the cell cycle. We counted absolute numbers of cytoplasmic and nascent transcripts (Zenklusen et al., 2008) and analyzed decay rates using a mathematical model without the use of transcriptional inhibitors, genetic mutants, or the need to normalize mRNA signal.
The use of a single-molecule mRNA decay measurement enabled identification of a novel regulatory pathway of mRNA decay that provides an additional level of cell cycle regulation. We determined that the half-life of SWI5 and CLB2 decreases more than 30 fold with the onset of prometaphase/metaphase. Furthermore, regulation of this mRNA decay is coordinated with their transcription and controlled by their promoter sequence, independent of the specific cis sequences located in the mRNA. By using morphological markers, we were able to determine that the cell cycle progression and the prometaphase/metaphase stability switch of SWI5 and CLB2 were coupled and regulated by the mitotic exit network kinases, Dbf2p and Dbf20p. Both kinases bind to SWI5 and CLB2 mRNAs, while Dbf2p is also enriched at their transcription sites. We propose a model whereby Dbf2p is first recruited by the promoter and then co-transcriptionally deposited onto the mRNA. Once in the cytoplasm, the mRNA associates with Dbf20p, and then waits for the appropriate cellular cues to initiate the decay process. Thus, for a subset of budding yeast mRNAs, promoter dependent activity directly influences how and when an mRNA will be degraded in the cytoplasm.
We measured SWI5 and CLB2 mRNAs decay rates in exponentially growing cells using the common approaches of qRT-PCR coupled with the transcriptional inhibitor thiolutin. A constitutively expressed ACT1 was expected to decay independently of the cell cycle phase with a single decay rate (t1/2 of 45 min) (Wang et al., 2002) and used as a control. Decay curves of SWI5, CLB2 and ACT1 (Fig. 1 A–C) were fitted to both an exponential decay with a single component (green line) and two components (red line) to identify which kinetic model best described their decay curves. A two component model detected a decay resistant SWI5 mRNA population (t1/2 > 90 min) and a rapidly decaying SWI5 mRNA population (t1/2 =3.0 min), whereas a single component model with a t1/2 = 6.9 min showed systematic deviations from the measured data. A two component model was not able to resolve multiple CLB2 decay populations and, similarly to ACT1 mRNA, fitted the decay data as well as a single component model. Here, CLB2 mRNA decayed with a single t1/2 = 3.7 min, while ACT1 mRNA decayed with a single t1/2 = 41.3 min, consistent with previously reported values (Wang et al., 2002).
To test whether SWI5 and CLB2 mRNAs decayed differently during the cell cycle, we synchronized cells in different cell cycle phases followed by thiolutin inhibition. In S phase and at G2/M border, SWI5 and CLB2 mRNAs were stable whereas in mitosis they decayed rapidly with an estimated t1/2 of ~3 min (Fig. 1 D–G, Fig. S1 A). ACT1 decayed independently of the cell cycle phase and, similarly to unsynchronized cells, turned over with a single t1/2 of ~30 min.
Kinetics of transcription inhibition by thiolutin was independent of the synchronization protocol (Fig. S1 B) and thus we could conclude that the stability of SWI5 and CLB2 mRNAs, but not of ACT1 mRNA, changed depending on the cell cycle phase. Two CLB2 decay populations, however, could only be detected when physically separated in time by cell culture synchronization. Therefore, normalization of mRNA decay signal, inhibition of transcription and use of population measurements obscured the behavior occurring in a fraction of cells, thus diminishing the sensitivity of the technique. We employed an approach that was both highly quantitative and minimally invasive to the cell’s physiology. We modified a FISH-based method that enabled us to quantify mRNA decay rates in individual, minimally perturbed cells with single mRNA sensitivity without the need for transcription inhibition, cell synchronization or normalization of mRNA signal.
We counted single transcripts in the cytoplasm in individual cells using single-cell, single-molecule FISH (Zenklusen et al., 2008). A mix of fluorescently labeled probes hybridizing along an mRNA was used (Fig. 2 A, red probes, Table S1), which strongly amplified the signal-to-noise ratio and detection sensitivity. Fluorescent transcripts were detected and counted using the algorithm Localize© ((Larson et al., 2005; Zenklusen et al., 2008), Fig. S2 A). Specific fluorescent signal was only detected in the presence of a target mRNA (Fig. S2 C,D). After cell segmentation, we obtained an absolute number of transcripts per cell, which obviated normalization of mRNA signal required for ensemble measurements and thus uncertainty associated with them.
FISH probes also annealed to the nascent chains whenever a cell actively transcribed a gene. We used a single probe targeted to the 5′-most end of the transcript labeled with a spectrally distinct fluorophore (Fig. 2 A, green probe, Fig. S2 B), and quantified the number of these probes annealed at the site of transcription (Femino et al., 1998; Zenklusen et al., 2008). This approach constitutes a direct measure of transcriptional activity in the cell, and the number of nascent chains reflects both the transcript initiation rate and the dwell time of a transcript as determined by all post-initiation processes, including elongation and termination (see below). By directly measuring this transcriptional output, transcription inhibition was no longer needed, which enabled us to measure kinetics of mRNA decay in chemically unperturbed cells.
To quantify changes in mRNA stability through mitotic division, we binned cells into cell cycle phases using morphological markers as indicators of cell cycle time (Brewer et al., 1984; Hartwell, 1974; Lord and Wheals, 1980) (Fig. 2 B). This approach provided temporal resolution without the need for cell synchronization. Four morphological markers were used: bud size (DIC), movement of the nucleus detected by DAPI, positioning of the spindle pole body indicated by CFP-tagged Spc42p, and localization of a GFP -tagged Whi5p. Whi5p is nuclear during telophase, cytokinesis and G1 phase of the cell cycle and cytoplasmic in all others (Bean et al., 2006), allowing differentiation between G1 phase and early S phase cells which have not yet formed a bud. Duplication time of the yeast culture was 90 +/− 8.5 minutes and the percent of cells in each phase was directly proportional to its length in minutes. For example, 18.3 % of cells were identified as late S phase cells which converted to 16.5 min (Fig. 2 B). Cell binning was highly reproducible among experiments using the same strain with a variability ≤10%. Thus, one can determine the relative time of progression through the cell cycle, allowing one to obtain dynamics of cell cycle gene expression from a population of fixed cells.
As anticipated for co-regulated genes, expression of SWI5 and CLB2 was similar and followed four discernable stages (Fig. 3 A–D). The first spanned the entire S phase, when transcription was infrequent and mRNA accumulation was modest. The second stage spanned G2 phase and prometaphase/metaphase, when transcription increased sharply and the bulk of mRNA synthesis occurred within 6.7 minutes. The onset of the third phase coincided with the onset of anaphase. Transcription of both genes ceased, and transcripts were degrading rapidly. In the last stage during G1 phase, the probability of expression of either of the genes fell below 5%.
Because transcription became inactive during mitosis, mRNA decay rates could be determined directly from their cytoplasmic mRNA profiles (Fig. 3 C,D blue line). During mitosis, SWI5 and CLB2 mRNAs decayed with a t1/2 of 2.1+/−0.8 min and a t1/2 of 1.8+/− 0.5 min respectively, based on the fit to the cellular RNA profiles. This rapid decay prevented carry-over of mRNAs into the new cell cycle (demarcated with gray boxes). Decay rates previously observed for SWI5 and CLB2 (Wang et al., 2002) were inconsistent with the data obtained by FISH; the slower decay rate would have contaminated the next cycle (Fig. 3 C, D green line).
To quantify CLB2 and SWI5 decay rates in the context of changing transcriptional activity during the cell cycle, we used mathematical modeling. The number of cytoplasmic mRNAs at any point of the cell cycle depends on the rates of their synthesis and decay. By measuring the total transcript level and the synthesis rate, we can determine the decay rate constant according to the differential equation:
where N is the number of transcripts in the cytoplasm (Fig. 2 A); t is time; m is the transcriptional activity of a gene measured by the number of nascent transcripts at a transcription site (Fig. 2 A); T is the dwell time of a nascent transcript (s); k is the degradation rate constant (s−1) (See Exp. Procedures). The solution to this differential equation is:
where N0 is the initial number of transcripts. The time t is determined from cell cycle markers as described above (Fig. 2B). The measured values are m and N determined as a function of t, and the fit parameters are k and T. For each gene a global non-linear least square fit to the expression profile was determined.
This model describes RNA levels as a balance of zero-order RNA synthesis and first-order decay. We assume first-order mRNA decay because it is the simplest model that describes our data and allows us to compare to bulk biochemical measurements of mRNA stability. In the simplest form, with a dwell time and an mRNA half-life that are invariant over the cell-cycle, both fit values are < 35s, resulting in RNA polymerase velocities and RNA lifetimes which are unphysical (Fig. S3 A–G). However, if we include the possibility of bimodal decay, as experimentally observed in Fig. 1, the model captures the essential features of SWI5 and CLB2 regulation over the cell-cycle. We kept the mitotic t1/2 fixed at 2.1 min for SWI5 and t1/2 at1.8 min for CLB2, as determined by the FISH measurements (Fig. 3 C,D, Fig. S3 C–G), and allowed the nascent transcript dwell time and the pre-mitotic decay to float. This fitting regime reached a global minimum with parameter values for dwell time of 66 s and a pre-mitotic t1/2 > 90 min for SWI5 mRNA (Fig. 3 E solid lines) and a dwell time of 63 s, and a pre-mitotic t1/2 of 66.1 min for CLB2 mRNA (Fig. 3 F solid lines). The dwell times consist of elongation and post-elongation processes, and if the termination time is estimated as 30 s (Zenklusen et al., 2008), the resulting polymerase velocity is approximately 50 bp/s, consistent with the polymerase velocity of 46 +/− 6.2 bp/s during S/G2/M (Larson et al., 2011). Additionally, the bimodal decay kinetics is invariant of nascent chain dwell times, as long as these remain within the physiological boundaries of T > 18 seconds (Fig. S3 C–G). Lastly, based on the best fit of our model, we determined that the switch in mRNA stability occurred during prometaphase/metaphase when transcription for both genes was at its peak, even though the rapid decay became apparent only with the onset of anaphase when transcription was shutting down (Fig. 3 A–F).
To further address the timing of destabilization, we modeled the data assuming that the switch in mRNA stability occurred with the onset of anaphase. This fitting regime also resulted in bimodal decay kinetics similar to the one described above, but with the higher divergence of the mathematical fit from the FISH data (data not shown). Based on the best fit criterion, this anaphase specific model was thus not considered in the analysis of SWI5 and CLB2 decay kinetics.
Several important conclusions are evident: 1) prior to mitosis, SWI5 and CLB2 transcripts were stable, allowing the cell to increase mRNA levels during active transcription; 2) during mitosis, when SWI5 and CLB2 transcription were shutting down, their transcripts decayed rapidly, preventing carry-over into the next cell cycle; 3) the switch from one state to the other occurred during prometaphase/metaphase (P/M) when SWI5 and CLB2 transcription reached its peak. These data suggest therefore that mRNA synthesis and decay were temporally coordinated for SWI5 and CLB2.
We sought to characterize the mechanisms that control the cell cycle dependent decay. Several regulatory elements could control degradation of an mRNA: its 5 ′ and 3′ UTRs, its ORF or its promoter. We thus replaced the SWI5 5′, 3′ UTRs and the promoter sequence with the corresponding regions of constitutively expressed ACT1 and DOA1 genes. We constructed cloning cassettes that allowed the integration of the constructs into the native SWI5 locus with a concurrent deletion of WT SWI5 copy (Fig. 4 A, Table S3). In wild type cells, ACT1 mRNA demonstrated a single decay rate of 41.5 min (Fig. 1 C–E) and accumulated transcripts throughout mitosis, unlike SWI5 (Fig. 4 B). WT DOA1 mRNA decayed with a t1/2 of 11.0 minutes and, similarly to ACT1, accumulated transcripts throughout mitosis (Fig. 4 C, Fig. S3 G, Fig. S4 A).
We tested the influence of SWI5 5′ and 3′ UTRs on the stability of SWI5 mRNA by replacing them with the 5′ and 3′ UTRs of the ACT1 gene (Fig. 4 D, Fig. S4 B). If binding of decay regulators to sequences in 5′ or 3′ UTR alone regulated SWI5 mRNA turnover, then this replacement should abolish cell cycle dependent mRNA decay and cause continuous accumulation of transcripts similar to WT ACT1 mRNA. However, this replacement had no effect on either the stability of the chimeric SWI5 mRNA or the prometaphase/metaphase-specific switch in mRNA stability. As in WT cells, chimeric SWI5 mRNAs were decay resistant prior to mitosis (t1/2 > 90 min) and decay sensitive during mitosis (t1/2 =2.4+/−1.3 min).
We then replaced only the SWI5 promoter with that of ACT1 while keeping the rest of the SWI5 mRNA intact (Fig. 4 E, Fig. S4 C). In this strain, SWI5 transcripts showed expression profiles similar to ACT1 with a single half-life of 19.7 minutes and continuous accumulation of transcripts throughout mitosis. Furthermore, when in addition to the SWI5 promoter we also changed its 5 ′ and 3′ UTR for that of ACT1 gene (Fig. S4 D), SWI5 transcripts decayed with a single decay rate of 18.7 minutes throughout the cell cycle.
In the reciprocal experiment, rapid decay during mitosis would bring ACT1 mRNA below the critical amount needed for the cell to survive. Therefore we used the non-essential DOA1 gene, which decays similarly to ACT1 (Fig. 4 C, Fig. S4 A), and expressed it using a SWI5 promoter (Fig. 4 F, Fig. S4 E). Under SWI5 promoter control, cell cycle dependent transcription and decay of SWI5 mRNA were recapitulated on DOA1 mRNA. DOA1 mRNAs were stable prior to mitosis with a t1/2 > 90 minute and decayed rapidly with a t1/2 of 4.9 +/−0.7 min during mitosis. Therefore, changes in the mRNA stability through the cell cycle were regulated entirely by SWI5 promoter and were independent of the specific cis mRNA sequences.
The same transcription factors that regulate SWI5 expression also regulate CLB2 expression (Darieva et al., 2006; Koranda et al., 2000; Spellman et al., 1998; Zhu et al., 2000) and thus it was likely that, as with SWI5, the cell cycle dependent decay of CLB2 was also controlled by its promoter. We thus replaced the promoter of CLB2 for that of ACT1, while keeping the rest of the CLB2 mRNA intact. When expressed from the ACT1 promoter, CLB2 transcripts turned over with a single mRNA half-life of 4.9 minutes and, unlike WT CLB2 mRNAs, accumulated continuously throughout mitosis (Fig. 5 A).
In the reciprocal experiment, a constitutively transcribed DOA1 was expressed from the CLB2 promoter and its mRNA stability measured. In this strain, the gene expression features of CLB2 were recapitulated on DOA1 with a slow mRNA turnover prior to mitosis (t1/2 of 14.7 min) and a rapid turnover during mitosis (t1/2 of 0.9 min) (Fig. 5 B). Unlike WT CLB2 however, the switch in the DOA1 mRNA stability occurred during telophase/cytokinesis. Due to the integration of cloning cassette into the CLB2 locus, this strain did not express the Clb2 cyclin. Consistent with the literature, this deletion resulted in an abnormal cell cycle and a delayed progression through mitosis (Fig. 5 B) (Fitch et al., 1992), which could have adverse effects on the decay process and the timing of the DOA1 mRNA stability switch.
These results demonstrate that the promoter sequence regulates the cell cycle dependent mRNA turnover of both the SWI5 and CLB2, independent of their cis mRNA sequences. mRNA stabilities measured for SWI5 and CLB2 when driven from the ACT1 promoter thus represent their “innate” abilities to resist decay. For SWI5 and CLB2 therefore, transcription and mRNA decay are co-dependent processes where the regulation of the first influences the outcome of the latter.
A bona fide regulator of SWI5 and CLB2 decay requires interaction with their transcription factors, the mRNA decay regulators and the cell cycle machinery to ensure coordination among the three. In the search of this trans-acting factor we made use of the Saccharomyces genome database. Because the regulation of SWI5 and CLB2 decay is promoter dependent, we asked whether any of their transcription factors physically interacted with a protein that in turn interacted with the mRNA decay regulators and the cell cycle regulators to provide coupling among the three processes (see Exp. Procedures). Dbf2p, a mitotic exit network (MEN) kinase, was the only protein that satisfied this criterion. It interacts with Cdc5p (Visintin and Amon, 2001), a SWI5 and CLB2 transcription factor and itself a MEN regulator (Darieva et al., 2006); it is a part of a larger CCR4-NOT complex (Liu et al., 1997), a major deadenylase of cytoplasmic mRNAs in yeast (Tucker et al., 2001); and Dbf2p is mitotically active to ensure telophase to G1 phase transition (Toyn and Johnston, 1994). Similarly, Dbf20p performs several Dbf2p functions and is synthetically lethal with Dbf2p (Toyn et al., 1991), so we assayed its role in regulation of SWI5 and CLB2 mRNA decay as well.
By using an RNA immunoprecipitation assay and mitosis synchronized cells to enrich for the SWI5 and CLB2 expression, we detected specific and significant binding of TAP-tagged Dbf2p to SWI5 and CLB2 mRNAs at the levels similar to Pab1p-TAP, but not to ACT1 and DOA1 mRNAs (Fig. 6 A,B). Significant Dbf20p-TAP binding was detected only with CLB2 mRNA. Furthermore, by using chromatin immunoprecipitation (ChIP) we detected significant enrichment of Dbf2p-TAP at SWI5 and CLB2 transcriptional units, which was also RNA dependent (Fig. 6 B,C,E, Fig. S5 A). Dbf2p-TAP binding was only detected in cells enriched in mitosis (red bars) and not in S phase (black bars). This result was anticipated since in S phase transcription of SWI5 and CLB2 was infrequent (Fig. 3 E,F red line) and thus the ChIP enrichment was not expected. Accordingly, the RNAPII ChIP in S phase cells was only marginally higher relative to background, particularly for SWI5 (Fig. S5 B). Co-transcriptional binding of Dbf20p-TAP to SWI5 and CLB2 mRNAs could not be detected (Fig. 6 B,D,E), indicating, that Dbf20p interacts with CLB2 after transcription is completed, possibly in the cytoplasm.
Finally, we measured SWI5 and CLB2 mRNA stabilities in the absence of Dbf2p and Dbf20p. Protein levels of either kinases do not fluctuate through the cell cycle (data not shown, (Visintin and Amon, 2001)) and hence we speculated that Dbf2p or Dbf20p could regulate either stable or unstable SWI5 and CLB2 mRNAs. Deletion of either of the kinases had no effect on the stability of ACT1 mRNA (Fig. S5 C,D), but greatly affected the stability of SWI5 and CLB2 mRNAs, particularly prior to mitosis (Fig. 6 F–I). Moreover, the regulation of mRNA stability by Dbf2p was independent of its kinase activity (Fig. S5 E,F). The mRNA half-lives determined for these two mRNAs using thiolutin and qRT-PCR were kinetically inconsistent with the FISH measurements (Fig. S5 G–L). These discrepancies are likely to have occurred due to toxic effects thiolutin exerts on the physiology of a cell and on the mRNA turnover (Jimenez et al., 1973; Pelechano and Perez-Ortin, 2008) thus artificially prolonging their mRNA stabilities.
Additionally, both deletions prolonged the G2 to T/C length of the cell cycle relative to the WT by two to four fold (Fig. S6 A–D), consistent with the literature (Liu et al., 1997). Thus, cells spent a longer time expressing SWI5 and CLB2 with transcriptional amplitudes similar to the WT cells but without excessive accumulation of transcripts (red circles in Fig. 6 F–I and Fig. 3 E,F). The measured reduced stabilities of SWI5 and CLB2 in ΔDBF2 and ΔDBF20 could not have been an artificial consequence of the redistribution of the transcripts over a longer cell cycle because the model accounted for the ongoing transcription. Thus, due to decreased stability of SWI5 and CLB2, cells were estimated to synthesize up to three times more mRNAs to reach the WT levels (Fig. S6 E–H).
These results imply that Dbf2p is recruited to SWI5 and CLB2 promoters, loaded onto their nascent chains co-transcriptionally, is exported with the mRNAs into the cytoplasm where along with Dbf20p regulates the timing of SWI5 and CLB2 decay (Fig. 7). Dbf2p and Dbf20p thus coordinate between SWI5 and CLB2 transcription and mRNA decay and communicate the cell cycle cues onto the decay machinery to initiate the decay process.
In this work, we developed a single-cell, single-molecule approach that enabled us to characterize the kinetics of mRNA decay with a high temporal resolution. This approach uncovered a unique, promoter dependent regulatory mechanism of mRNA decay that could be employed by a variety of cell cycle regulated genes. We further identified two novel regulators of mRNA decay, Dbf2p and Dbf20p, each with distinct functions in regulating SWI5 and CLB2 stability. A particular advantage of our approach was that, unlike traditional techniques used to quantify mRNA turnover, we were able to measure transcription and decay concurrently. This enabled us to determine that SWI5 and CLB2 transcription and decay were temporally coordinated through the cell cycle. Furthermore, deletion of Dbf2p and Dbf20p resulted in destabilization of SWI5 and CLB2 mRNAs and also reduced the efficiency of coordination between transcription and decay. Thus, cells spent longer time transcribing SWI5 and CLB2 without excessive accumulation of mRNAs. To compensate for the increased decay, cells were estimated to synthesize up to four times more mRNAs to reach the WT levels. Our measurements reveal, therefore, that balancing infrequent transcription with mRNA stability is necessary for effective transcript build-up, while transcription shut-down during mitosis and rapid decay prevent carry-over of mRNAs into the next cell cycle.
Achieving specificity of mRNA decay through a promoter sequence and not a specific cis mRNA sequence, as shown for SWI5 and CLB2, is a unique attribute. In order to maintain coordination between transcription and cytoplasmic decay, only the promoter sequence needs to be conserved while the mRNA sequence can vary independently without disrupting regulation of either of the processes. If then multiple genes share promoter sequences, the entire expression process can be coordinated as a group. For example, SWI5 and CLB2 share promoter sequences with 33 other genes in the CLB2 cluster involved in the G2/M transition (Koranda et al., 2000; Spellman et al., 1998; Zhu et al., 2000) and may be similarly regulated. Several of these genes are toxic when overexpressed (Niu et al., 2008; Sopko et al., 2006), indicating that timing of their expression during the cell cycle is restricted. Such coordination would ensure that all mRNAs in a group would oscillate as one entity, ensuring sharp transitions between cell cycle phases. It is possible that the promoter-dependent coordination between transcription and mRNA decay could be employed by cell-cycle cluster genes other than the CLB2 (there are eight in budding yeast (Spellman et al., 1998)). Interestingly, several transcripts in budding yeast couple the regulation of their stability with the transcriptional activity of RNAPII through polymerase-interacting subunits Rbp4/7 (Goler-Baron et al., 2008). This coupling phenomenon is employed by approximately ten percent of the genes in Saccharomyces cerevisiae and was shown to be preserved through evolution (Dori-Bachash et al., 2011). Thus, intriguingly, promoter-dependent regulation of mRNA stability could be a common strategy of control of mRNAs turnover in yeast and possibly in a variety of eukaryotic cells.
We propose a model where the regulation of cytoplasmic SWI5 and CLB2 mRNA decay begins concurrently with their transcription (Fig. 7). To provide specificity of decay, independent of specific cis mRNAs sequences, the decay regulator must be recruited by SWI5 and CLB2 transcription factors to their promoters and deposited onto the mRNA during transcription. Possibly the promoter recruits factors that influence a specific mRNA feature, such as the cap structure, the poly(A) tail or their associated proteins. We identified Dbf2p kinase as a regulator of SWI5 and CLB2 decay and found it associated with their mRNAs during transcription. SWI5 and CLB2 are additionally stabilized by Dbf20p. Unlike Dbf2p, Dbf20p does not bind to SWI5 and CLB2 co-transcriptionally and likely associates with mRNAs in the cytoplasm. Our data indicates therefore that despite being redundant, Dbf2p and Dbf20p have distinct functions in regulation of SWI5 and CLB2 mRNA stability, indicating that their roles in the decay process are complex and could involve multiple regulators.
How Dbf2p becomes recruited to the promoters of SWI5 and CLB2 and how the two kinases interact with the mRNAs is not clear. Interaction of Dbf2p with Cdc5p (Visintin and Amon, 2001), a SWI5 and CLB2 transcription factor (Darieva et al., 2006) suggests a possible mechanism. Additionally, the mechanism whereby Dbf2p and Dbf20p regulate mRNA stability is also unknown. This regulation is independent of Dbf2p (and presumably Dbf20p) kinase activity, which is triggered shortly after metaphase to anaphase transition to promote progression from telophase to G1 phase (Toyn and Johnston, 1994). These results are consistent with our findings that Dbf2p and Dbf20p stabilize SWI5 and CLB2 mRNAs prior to mitosis when their kinase activity is low (Toyn and Johnston, 1994). Because the prometaphase/metaphase and anaphase are separated by only a couple of minutes it is possible that insufficient time resolution during mitosis obscured precisely when the mRNA stability switch occurs. Nevertheless, our data indicates that Dbf2p and Dbf20p have two biologically distinct and mutually independent functions: a novel one involved in regulation of mRNA stability described here and a better understood one involved in the regulation of completion of mitosis as MEN regulators (Mah et al., 2001).
Additionally, how Dbf2p and Dbf20p relay cell cycle signals onto the mRNA decay machinery to initiate decay remains to be determined. Dbf2p is dephosphorylated during mitosis (Toyn and Johnston, 1994), and we speculate that this dephosphorylation event could act as a cell cycle signal thereby synchronizing mRNA degradation and mitotic division. Intriguingly, association of Dbf2p with the CCR4-NOT complex suggests that regulation of decay could be manifested through the regulation of deadenylation as determined for tristetraprolin protein TTP. Dephosphorylation of TTP controls if and when CCR4-NOT complex is able to gain access to the mRNA to initiate decay (Clement et al., 2011; Sandler et al., 2011). Similarly to TTP, Dbf2p and Dbf20p might regulate accessibility of CCR4-NOT complex to the SWI5 and CLB2 mRNAs in a dephosphorylation-dependent but kinase activity-independent manner.
Here, we show that the fate of the SWI5 and CLB2 mRNA is determined co-transcriptionally at their birth. Thus, the decay marker assembles on the mRNA, is exported with it into the cytoplasm, priming the mRNAs for immediate decay once a cell cycle signal arrives. Furthermore, in budding yeast, transcriptional activity can directly determine how an mRNA will localize, translate and degrade in the cytoplasm (Harel-Sharvit et al., 2010; Shen et al., 2010). Thus, we hypothesize that a subset of yeast mRNAs could become “fully functionally configured” during their synthesis. These mRNAs could exit the nucleus equipped with the regulatory proteins that would define their translation, localization and decay, which would then be “shed away” from an mRNA in a step-by-step manner after each completed step (Trcek and Singer, 2010). This model of mRNP formation is quite different from the one generally assumed for an mRNA, where proteins that regulate different steps in an mRNA life path interact with an mRNA only when their function is needed (Balagopal and Parker, 2009). Our study may thus have far-reaching implications that will serve as a platform for the analysis of mRNA decay and proteins that regulate it in a variety of mRNAs and organisms.
Table S3 and Suppl. Exp. Procedures lists yeast strains used, their synchronization and growth conditions.
Per gene, three to seven probes were used, each labeled with > 90% labeling efficiency (Table S1). Design, synthesis and labeling of probes was performed as described previously (Femino et al., 2003; Zenklusen et al., 2008).
ACT1 mRNA was highly expressed and therefore reliable counting of single transcripts in a maximal projection as performed for SWI5 and CLB2 was not possible. Instead, images were sum projected and total fluorescent intensity of ACT1 FISH signal for each cell measured and presented as an average. The summed fluorescent values were corrected for the autofluorescent cellular background of the same cellular size from the control cells not hybridized with ACT1 probes. The control cells were subjected to the same hybridization procedure and imaged as ACT1 FISH cells only without the ACT1 probes.
The number of transcripts measured in a particular phase of the cell cycle is the time-integrated average of the time-dependent solution (Eq. 2) divided by the length of that particular cell-cycle phase:
where brackets denote the ensemble average over the population of cells in a particular cell cycle phase; Tc is the duration of that phase; m, T, k are defined previously (Eq. 2) as the number of nascent chains, the dwell time and the decay rate, respectively. The initial number of transcripts N0 is determined by the number of transcripts present at the end of the previous cell cycle stage:
where i designates the cell-cycle phase. Thus, the initial number of transcripts N0i is determined from the time-dependent solution N(t) (Eqn. 2) at a time Tc corresponding to the length of the previous cell-cycle phase, where the kinetic values mi-1, ki-l, are also those of the previous cell-cycle phase. The dwell time (T) of a nascent chain at the gene is determined by the parameters v (RNAPII velocity) and l (transcript length) (see Table S3, S4). Equations 3 and 4 were used to model the data in Fig. 3–6 and their supplemental data.
Table S4 summarizes the parameters used to model the FISH data. For WT SWI5, CLB2, SWI5 with ACT1 5′ and 3′ UTRs and for DOA1 expressed from the SWI5 promoter, the mitotic decay was measured by fitting their cytoplasmic mRNA abundances after anaphase onset to an exponential decay with a single component. A slow decay was determined by calculating a global non-linear least square fit to the N and m with two floating parameters (T and a pre-mitotic decay rate) while the mitotic decay (from P/M to T/C) rate measured by FISH was fixed.
For WT DOA1, SWI5 expressed from the ACT1 promoter, SWI5 with ACT1 5′ and 3′ UTRs expressed from the ACT1 promoter and CLB2 expressed from the ACT1 promoter, the data with one free parameter (a single k) were modeled. The velocity v of the RNAPII of 33 bp/s was assumed (Mason and Struhl, 2005) to obtain the dwell time T of 68 s (WT DOA1), 72 s (SWI5 with an ACT1 promoter), 70 s (SWI5 with ACT1 5′ and 3′ UTRs expressed from the ACT1 promoter) and 66 s (CLB2 with an ACT1 promoter).
For ΔDbf2 and ΔDbf20 deletions, for DOA1 mRNA expressed from a CLB2 promoter and for the Dbf2p Kinase Dead experiment, the mitotic decay could not be determined directly from their cytoplasmic mRNA profiles, because in these strains the mitotic phases were extended two to three times relative to WT with and the addition of new mRNAs due to ongoing transcription was not negligible. Here, the data was modeled with the fixed T of 66 s for SWI5, 63 s for CLB2 and 77 s for DOA1 determined for the WT SWI5 and CLB2 respectively, while the pre-mitotic and the mitotic decay rates were free parameters.
A bona fide regulator of SWI5 and CLB2 decay requires interaction with their transcription factors, the mRNA decay regulators and the cell cycle machinery to ensure coordination among the three. Cell cycle dependent transcription of SWI5 and CLB2 is regulated by four transcription factors, Ndd1p, Fkh2p, Mcm1p and Cdc5p and their promoter binding positions have been determined (Darieva et al., 2006; Koranda et al., 2000; Spellman et al., 1998; Zhu et al., 2000). We reasoned that since the stability of SWI5 and CLB2 is promoter specified, then the mRNA decay regulator we were searching for thus has to be recruited to SWI5 and CLB2 promoters by one of their transcription factors to ensure specificity of decay. This regulator in turn has to interact or be a part of the mRNA decay machinery and the cell cycle progression machinery to further enable the coordination of decay through mitotic division.
In the search of this trans-acting factor we made use of the Saccharomyces genome database. Ndd1p, Fkh2p, Mcm1p and Cdc5p each uniquely interacted with five, 14, 20 and 97 proteins respectively. Dbf2p, a mitotic exit network (MEN) kinase, was the only protein that satisfied our criterion; it interacts with Cdc5p (Visintin and Amon, 2001), a SWI5 and CLB2 transcription factor and itself a MEN regulator (Darieva et al., 2006), it is furthermore a part of a larger, 1.9 MDa CCR4-NOT complex (Liu et al., 1997), a major deadenylase of cytoplasmic mRNAs in yeast (Tucker et al., 2001) and is mitotically active to ensure telophase to G1 phase transition (Toyn and Johnston, 1994). Dbf2p interacts with four out of nine proteins of the CCR4-NOT complex; with Ccr4p, the catalytic subunit of CCR4-NOT complex with deadenylase activity (Tucker et al., 2002), with Pop2p, Caf40p and Caf36p, non-catalytic subunits of CCR4-NOT complex, with Caf4p, a CCR4-NOT associated protein and Cdc33p and Cdc20p, a 5′ cap binding protein and a cap associated protein. Dbf2p co-purifies with all components of the complex itself and co-immunoprecipitates with the Ccr4p and Pop2p proteins (Liu et al., 1997). Additionally, ΔDBF2 results in similar phenotypes and transcriptional defects to those observed in ΔCCR4 and ΔPOP2. Conversely, ΔCCR4 and ΔPOP2 affected mitotic cell cycle progression similar to that observed for ΔDBF2 indicating that Ccr4p, Pop2p and Dbf2p all participate in regulating gene expression and cell cycle progression during late mitosis (Liu et al., 1997).
Dbf2p is synthetically lethal with Dbf20p, not known to interact with the CCR4-NOT complex or Cdc5p like Dbf2p, but performs several Dbf2p functions (Toyn et al., 1991), so we assayed its role in regulation of SWI5 and CLB2 mRNA decay as well.
Apart from Dbf2p, SWI5 and CLB2 transcription factors displayed other interactions, but only either with the major mRNA decay regulators or cell cycle progression regulators, not both, thus making them unsuitable candidates. For example, Mcm1p interacted with Arg81 and Cdc5p interacted with Cse4p that in turn interacted with Dcp2p, a catalytic subunit of the Dcp1p-Dcp2p decapping enzyme complex. Cdc5p also interacted with Mcd1p that in turn interacted with Not5p, a subunit of the CCR4-NOT complex and with Nop13p and Pds5p that in turn interacted with Xrn1p, a 5'-3' exonuclease and component of cytoplasmic processing (P) bodies involved in mRNA decay.
Finally, apart from its role as a transcription factor, Cdc5p is mostly known as a MEN regulator in promoting transition of cells from telophase into G1 phase (Toyn and Johnston, 1994). It physically interacts with several MEN regulators, for example with Dbf2p. Unlike Dbf2p however, none of these regulators in turn interact with the mRNA decay factors, making these proteins unsuitable candidates involved in the regulation of SWI5 and CLB2 mRNA stability.
Unless cited, the protein descriptions were obtained from the Saccharomyces genome database.
45 ml of cells were grown in YPD until OD600 ~0.35. Cells were synchronized in S or M phase with HU (see Suppl. Exp. Procedures). The ChIP was performed as described in ((Moldon et al., 2008), Table S5). For RNase ChIP in mitotic cells, crosslinked extracts were treated with DNase-free RNase (50 µg/ml, Roche) for 15 min at 37°C prior to sonication. The RNA-IP was performed as described in (Gilbert et al., 2004).
See Table S6 and the accompanying text.
We would like to thank Dr. Jeffrey Chao and Dr. Kevin Czaplinski for helpful discussion of the project and Xiuhua Meng for her help with cloning. We thank Drs. Ian Willis, Michael Keogh and Angelika Amon for sharing ΔSWI5 (TT012), Pab1p-TAP (TT065) strains and Dbf2p-Kinase Dead plasmid with us, respectively. This work was supported by an EMBO fellowship awarded to AM, GM57829 to CCQ and GM57071 to RHS.
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.