Introduction

The post-transcriptional level of regulation comprises many steps from synthesis of the primary transcript to the final degradation of the mRNA molecule: 5′-capping, splicing, poly(A)-site cleavage and polyadenylation, RNA editing, nonsense-mediated decay, nuclear export, localisation in cytoplasmic compartments, initiation and elongation of translation, and the different steps in mRNA decay.

In the last years, it has become clear that certain modes of post-transcriptional control may be quite widespread and can have far-reaching consequences. Thus, alternative splicing is being put forward as an important source of protein variety for multicellular organisms, thus acting as a qualitative expression control. In many higher plants and animals, the required numbers of different proteins are sometimes perceived to greatly exceed the number of recognised genes (Roberts and Smith 2002). To what extent alternative splicing is modulated by signalling pathways remains comparatively unexplored. Regulation through small RNAs (RNA interference; RNAi) can affect translation and stability of mRNA, but can also exert a more long-lasting effect on transcription through establishment of heterochromatin.

The emphasis in this review is on quantitative control mediated through signalling pathways of the later, cytoplasmic, steps in post-transcriptional control: regulation of translational initiation and elongation, through mechanisms including upstream open reading frames, and mRNA turnover. Since post-transcriptional regulation acts on pre-existing mRNAs and thus is inherently faster than regulation of transcriptional initiation, it may be more relevant in certain situations to view events on the post-transcriptional level after, e.g. hormone stimulation as earlier in the causative chain than up-regulation of transcription (Prendergast 2003).

There are well-documented cases of post-transcriptional control in organism development and as a consequence of particular circumstances. Thus, extensive post-transcriptional regulation is of necessity in neurons, with an extremely long distance to the nucleus (Klann et al. 2004). Those cases will not be considered here as the intent is to highlight short-term effects in response to abiotic or hormonal stimuli.

Post-transcriptional control

Most of the control exerted on translation is believed to act on the initiation step. For translation to initiate, the amino acid-charged initiator tRNA must first associate with the small subunit of the ribosome. Then, the mRNA attaches and the start codon is recognised. Finally, the large ribosomal subunit joins. The individual steps of this process are facilitated by translation initiation factors (abbreviated eIFs in eukaryotes). As initiation of translation requires interactions between the ribosome and both ends of the transcript, many eIFs bind to the 5′ or the 3′ ends of mRNAs. Much of the known regulation of translation through signalling pathways converges at eIF2α (Fig. 1a). The initiation factor eIF2 in its GTP-bound form together with charged initiator tRNA (ternary complex) is brought to the 40S subunit of the ribosome, to form a preinitiation complex. Phosphorylation of eIF2α-GDP (in mammals by either of the four kinases GCN2, PERK, HRI or PKR; in budding yeast solely by Gcn2) converts it to a high-affinity inhibitor of eIF2B, its own GDP/GTP exchange factor, preventing recharging of eIF2α with GTP. The overall effect is a down-regulation of translation initiation. Another central mechanism for regulation of the initiation step occurs through eIF4E-binding proteins (4E-BPs; see Fig. 1b). These are a family of small proteins that share a conserved binding site for eIF4E. By competing with eIF4G (which has the same binding site) for binding to eIF4E, a 4E-BP can prevent formation of the cap-binding eIF4E-eIF4G-eIF4A complex, collectively called eIF4F, which is required for all cap-dependent translation. Binding of a 4E-BP to eIF4E is modulated by phosphorylation through a multitude of signalling pathways, after activation by, e.g. growth factors or cytokines. Hyperphosphorylation of 4E-BP prevents binding to eIF4E, ultimately leading to activation of translation. For reviews of general translation and control of translation initiation, see Dever 2002 and Preiss and Hentze 2003.

Fig. 1
figure 1

Overview of mechanisms for regulation of translation initiation. a Inhibition of translation initiation through stress-induced phosphorylation of eIF2α by Gcn2. The phosphorylated form of eIF2α inhibits eIF2B, preventing recharging of eIF2α with GTP, which leads to decreased initiation. b Stimulation of translation initiation through growth stimuli-induced phosphorylation of 4E-BP. The 4E-BPs use the same binding surface on eIF4E as eIF4G. The phosphorylated forms of 4E-BP are unable to bind. Adapted from Dever 2002 and Gallie 1998

Full size image

Two major pathways of mRNA degradation exist in eukaryotes. In both cases, shortening of the poly(A) tail is the first, time-limiting, step. Three distinct protein complexes (the Pan2/Pan3, or PAN complex; poly(A)-specific exonuclease, PARN; and the Ccr4/Pop2 complex) govern this deadenylation. After deadenylation, degradation can occur from the 3′ to the 5′ end by the RNase-containing exosome complex. In an independent pathway, deadenylation is followed by removal of the 7-methyl-guanosine cap of mRNAs and then proceeds in the 5′–3′ direction. The mechanisms of mRNA turnover have been reviewed recently (Meyer et al. 2004).

There is a competition between ribosomes and ribonucleases for binding to mRNA, so actively translated mRNAs are generally protected from translation. Insertion of elements in the 5′-UTR that interfere with efficient translation will destabilise the mRNA (Muhlrad et al. 1995). PolyA-tail shortening proteins generally compete with translation factors and the ribosome for access to mRNA, and so highly translated mRNAs are degraded more slowly than weakly translated ones (Prieto et al. 2000). Stalling of ribosomes at internal sites in an mRNA will also induce endonucleolytic cleavage and degradation (Doma and Parker 2006). In general, therefore, for a given mRNA species we should expect a negative correlation between translational activity and degradation.

Regulation through upstream open reading frames

An additional mechanism for translational control operates through short open reading frames (upstream open reading frames; uORFs) present in the 5′-UTR of certain genes. These can interfere with translation of the main ORF by preventing reinitiation after translation has been terminated at the end of the uORF. The paradigm case for uORF regulation is Saccharomyces cerevisiae GCN4. There, translation of uORF4 located close to the start codon of the main ORF precludes translation of the main ORF. Translation of uORF1, located further upstream, can result in reinitiation downstream of it. The different probabilities of reinitiation downstream of uORF1 and uORF4 stem from the shortness of uORF1 and A/U-rich sequences immediately downstream of it, and uORF4 being longer and having G/C-rich sequences downstream. If reinitiation occurs at uORF4, translation of the main ORF will be prevented. If scanning ribosomes can ignore the uORF4 start codon and continue scanning downstream of uORF4, however, the GCN4 main ORF will be translated and Gcn4, a transcription factor regulating many genes required in amino acid metabolism, will be produced (Abastado et al. 1991; Hinnebusch 1996). The nutrition status of the cell modulates GCN4 translation through the level of ternary complex. Under conditions of high nutrient availability, the concentration of ternary complex is high. This leads to a high frequency of reinitiation, and so uORF4 will be translated, but not GCN4 itself. Starvation activates Gcn2 to phosphorylate eIF2α, reducing ternary complex levels, and so the reinitiation frequency decreases. This means scanning will resume downstream of uORF4, and so translation of Gcn4 increases (Dever et al. 1992).

Regulation through uORFs may be more widespread than believed previously. Genome comparisons between man and mouse (Crowe et al. 2006), Aspergillus species (Galagan et al. 2005), and Saccharomyces species (Zhang and Dietrich 2005) indicate that many uORFs are conserved in evolution with respect to sequence and position.

Localised degradation and storage of mRNA

There is evidence that in undisturbed cells, mRNA decay occurs in discrete cytoplasmic foci containing many components involved in mRNA degradation such as decapping enzymes, exonucleases, and mRNA degradation intermediates. Such structures have alternately been called processing bodies (P bodies), cytoplasmic foci, or GW bodies for their content of the glycine–tryptophan repeat-rich protein GW182 (Cougot et al. 2004; Sheth and Parker 2003; Yang et al. 2004). Under conditions of severe stress (oxidative, metal, and heat), eukaryotic cells in addition accumulate cytoplasmic particles called “stress granules”, which have been shown to contain mRNA as well as ribosomal components, translation initiation factors and specialised proteins required for their structural integrity. Stress granules were first described in plant cells (Nover et al. 1983) and later in mammalian cells (Kedersha et al. 1999) as well as in the fission yeast (Dunand-Sauthier et al. 2002). So far they have not been unequivocally demonstrated to exist in the budding yeast, however. Formation of stress granules depends on interaction domains of the proteins TIA and TIA-R in a process that has been compared to prion formation (Gilks et al. 2004; Kedersha et al. 2000). mRNAs contained in stress granules or P bodies are translationally silent.

A direct connection between polysomes on one hand, and stress granules and cytoplasmic foci on the other, becomes obvious through the interesting observations that drugs such as emetine or cycloheximide, translation inhibitors that also block disassembly of polysomes, simultaneously inhibit formation of stress granules under arsenite stress, or accumulation of cytoplasmic foci. By contrast, application of puromycin, which induces premature termination of translation and dissociation of polysomes, causes stress granule formation even without external stress factors (Cougot et al. 2004; Kedersha et al. 2000). This suggests a dynamic relationship between mRNA at actively translating ribosomes and stress granules. When the cell encounters strongly adverse conditions, one of its first actions is to strongly down-regulate the translational machinery. This serves the dual purpose of conserving energy—protein synthesis represents a major part of the cell’s energy expenditure—and of facilitating rapid redirection of synthesis resources from gene products required for rapid growth under near-optimal conditions to products protecting the cell from stress. The amount of polysomes in such cells rapidly decreases, and in cases where stress conditions are severe, stress granules appear in the cytoplasm. It is natural to presume that mRNA and ribosomal components from disassembled polysomes appear in stress granules, which are characterised by a rapid exchange of RNA and proteins with the environment.

A model for the relationship between stress granules and P bodies has been proposed (Kedersha et al. 2005): mRNAs are first sorted from polysomes to stress granules under conditions of severe stress. They can then be recruited back to polysomes, or alternatively, be transferred to P bodies for degradation. P bodies can also form in the absence of stress granules, however (Kedersha et al. 2005). Both stress granules and P bodies contain components of the 5′–3′ mRNA degradation pathway (Kedersha et al. 2005; Sheth and Parker 2003; Yang et al. 2004); a composition difference between them is that translation initiation factors are mainly found in stress granules but not in P bodies (Kedersha et al. 2005). It should be noted that evidence also exists for recruitment of mRNA from P bodies to polysomes in budding yeast when translation resumes (Brengues et al. 2005).

Signalling through phosphorylation of eIF2α fundamentally affects formation of stress granules. Expression of a mutated allele mimicking the phosphorylated form of eIF2α causes stress granules to form even in the absence of external stimuli (Kedersha et al. 1999; McEwen et al. 2005), whereas an allele expressing a non-phosphorylatable form of eIF2α blocks formation of stress granules (Kedersha et al. 1999).

Techniques to study post-transcriptional regulation

Translational activity

For an individual protein, it is possible to determine the formation rate through short term labelling in vivo with radioactive amino acid followed by immunoprecipitation. The workhorses of proteomics, 2D gel electrophoresis (2D-PAGE) and mass spectrometry, determine steady-state levels of proteins, which in turn are a function of the individual production and degradation rates. Pulse-labelling techniques can be used to estimate individual protein synthesis rates in 2D-PAGE, but are limited to the more abundant proteins. To measure degradation rates on a global scale, Belle et al. 2006 employed an exhaustive collection of S. cerevisiae strains expressing TAP-tagged proteins, where individual decay rates were measured after inhibition of protein synthesis with cycloheximide.

Polysomal association

When in the elongation phase of translation, mRNAs are associated with typically several ribosomes, an arrangement called a polysome. Because of their high molecular weight and abundance, polysomes can be easily separated from other fractions. This can be done both analytically and preparatively, by sedimentation ultracentrifugation of crude lysates, treated with cycloheximide to prevent ribosomes from dissociating from mRNAs (Foyt et al. 1991; Marcus et al. 1967). After separation, the contents of the centrifugation tube can be collected from one end of the gradient. By online UV monitoring, one obtains a polysomal profile (Fig. 2), which itself can be used to diagnose many aspects of the overall translational state of the cell. The amount of polysomes reflects the overall level of translational activity in the cell. Further, the gradient can be fractionated and the fractions analysed for their content of individual mRNAs by hybridisation. Northern blotting can be used to monitor the precise distribution between the different polysomal, monosomal, and submonosomal peaks for a particular mRNA. It is also feasible to use the fractions from a polysome gradient (individual peaks or, e.g. pools of polysomes) for hybridisation to DNA arrays. By comparing the intensity patterns obtained from polysomal RNA with those from total or subpolysomal RNA, we can deduce the translational efficiency (as measured by the specific polysomal association) for any mRNA.

Fig. 2
figure 2

Polysomal profiles from S. cerevisiae cells with or without exposure to oxidative stress. Total cell lysates were prepared in the presence of cycloheximide to preserve ribosome/mRNA complexes, and separated on sucrose gradients by ultracentrifugation. Absorbance at 254 nm was recorded online while collecting the gradients; higher mass complexes are located to the right in the diagram. 40S, 60S, and monosomes represent the small, the large ribosomal subunit, and the 80S “monosomal” peak, respectively. The numbers above peaks in the polysomal range represents the number of mRNA-associated ribosomes. Adapted from Swaminathan et al. 2006

Full size image

It is logical to assume that mRNAs associated with two or more ribosomes are actually being actively translated, although it is conceivable that under conditions of cellular stress or nutrient deprivation a fraction of them could represent cases where translation has stopped, but dissociation of the complexes has not yet taken place. The interpretation of the monosome peak is more ambiguous. Beside mRNAs where initiation has taken place but elongation not yet started, this fraction also contains ribosomes in the elongation phase attached to mRNAs that are too short to house more than one ribosome at a time, and mRNAs where the translation rate is so low that they are on average occupied by only one ribosome. To bypass this uncertainty, the majority of analyses exclude the monosome peak from the fraction of actively translated mRNAs and consider only the polysomal peaks.

mRNA decay

In vivo labelling methods

It is possible to estimate the half-life of individual mRNAs in intact cells without interfering with either general transcription or the transcription rate of the gene of interest. In “approach to steady-state” labelling (Kim and Warner 1983), radioactive RNA precursors are added to living cells for various time lengths, followed by extraction of total RNA and hybridisation to immobilised probes. The rate by which the specific radioactivity comes closer to that obtained at extended labelling times (steady-state) is an indirect measure of the turnover time of the pool of the mRNA species in question. The advantage of this method is that the cells are left undisturbed, including the transcription rate of the gene under study. Drawbacks include the relatively low sensitivity, limiting this approach to highly expressed genes, and a low temporal resolution.

Promoter switch-off

Putting a gene of interest under control of a switchable promoter allows one to follow the fate of an individual mRNA species. If transcription can be rapidly turned off, then the decay of the mRNA species in question can be monitored. Popular promoters with this property that are used in yeast include Schizosaccharomyces pombe nmt1, S. cerevisiae GAL1, and variations of the heterologous Tet system, where transcription is quickly terminated by addition of thiamine, glucose, or a tetracycline-type drug, respectively. Obvious disadvantages of this approach are the workload involved in creation of the regulatable constructs, and that only one mRNA species can be measured in an individual strain. In the case of glucose, genes under regulation of carbon source will be more difficult to study using this regime.

Arrest of RNA polymerase II

Another avenue is to abolish all transcription by RNA polymerase II (Pol II). This allows simultaneous measuring of individual half-lives of all mRNAs in the cell under the same conditions. Clearly, a cell where mRNAs are no longer being made is a dying cell. Interpretation of this type of experiments generally relies on the assumption that the integrity of the cell is maintained, including the state of mRNA degradation machineries, and so one always has to keep reservations in mind about possible artefacts arising because of this major disturbance of the state of the cell.

The first way to halt transcription is the use of conditional mutants defective in Pol II components. Temperature-sensitive Pol II mutants such as rpb1-1 (Nonet et al. 1987) are readily available in yeast. Temperature upshift causes Pol II transcription to arrest within a few minutes in such a mutant. Studies using temperature-sensitive Pol II mutants have provided a large body of valuable insights into the mechanisms of mRNA decay. They suffer, obviously, from the limitation that controls have to be provided, as far as possible, to account for the effects of the temperature shift itself.

The other option is to arrest Pol II by chemical inhibitors, side-stepping the requirement for a temperature shift (Caponigro and Parker 1996 and references therein). The most commonly used drug is 1-10-phenantroline, a zinc chelator. Although zinc is a co-factor of many enzymes, the effect of intracellular zinc exhaustion on RNA polymerases, particularly Pol II, is immediate and extensive. As a result, cellular mRNA production virtually ceases shortly after addition of 1-10-phenantroline. Side effects of the chemical inhibitor are an issue to be considered. Beside inhibitory effects on other enzymes than the targeted one (such as loss of activity of zinc-requiring proteins in the example of 1-10-phenantroline), we have to consider general stress responses (Causton et al. 2001; Gasch et al. 2000) as well as compensatory mechanisms elicited by the drug. This can be seen, e.g. in 1-10-phenantroline treatment as up-regulation of several genes required for zinc import and metabolism. A systematic comparison of the readout in the form of mRNA decay rates has been performed of different Pol II inhibitors (1-10-phenantroline, thiolutin, 6-aza-uracil, ethidium bromide, and cordycepin) side by side with rbs1-1 mutants (Grigull et al. 2004). It was found that thiolutin and 1-10-phenantroline produced patterns that were most similar to that of rbs1-1 mutants. Overall, there was a high degree of overlap between all five different inhibitors and rbs1-1 mutants, as expected. However, certain subgroups of genes display aberrant, inhibitor-specific, effects.

High-throughput methods

Two of the techniques described above, Pol II inactivation and polysomal association, lend themselves well to global studies using DNA arrays. Polysomal analysis measures physical association with ribosomes, and is thus only an indirect estimation of translational activity. Because of its appropriateness for large-scale studies, however, it can be used to connect proteomic and transcript array data. Global technologies for study of translational activity have recently been surveyed (Beilharz and Preiss 2004). As already pointed out, it is also possible to measure mRNA decay rates on a global scale using DNA arrays (Fig. 3). A technical issue in such experiments is the need for an external or internal standard that is assumed not to change its abundance with time. An avenue sometimes taken is to use groups of RNAs known to be stable, and measure mRNA decay rates relative to those.

Fig. 3
figure 3

Determination of mRNA decay rates using microarray analysis. Synthesis of mRNA is arrested using an inhibitor of RNA polymerase II such as 1-10-phenantroline. RNA samples taken from wild-type or mutant cells before phenantroline addition are hybridised to arrays together with samples taken at various times after addition. Decay rates of individual mRNAs are calculated from differences in relative hybridisation intensity. In the bottom diagram, filled circles represent an mRNA under stability control of the pathway affected in the mutant

Full size image

To analyse post-transcriptional regulation requires more detailed information about the fine structure of RNA species expressed from a gene. Identification of genes with a potential for post-transcriptional regulation should be greatly enhanced by the novel high-density arrays that have been used to fine-map the yeast transcriptome (David et al. 2006), yielding the exact ends of each transcript, as well as the presence of anti-sense transcripts.

A biochemical global technique complementing estimates of transcript stability and polysome association is isolation of the mRNA bound to an affinity-tag labelled protein, followed by DNA array analysis of the mRNA population compositions. This has allowed functional profiling of RNA-stabilising proteins and translation initiation factors (Duttagupta et al. 2005; Gerber et al. 2004, 2006; Zhou et al. 2005). For example, the homologue of TIA/TIA-R in yeast, Pub1, has been shown by array analysis of the associated mRNA population to selectively bind about 5% of all mRNA species. At the same time, about 10% of all transcripts display a reduced half-life in a pub1 mutant. The mRNA set which was found to be both physically associated with Pub1 and to require this protein for stability encoded mainly proteins for ribosome biogenesis and intermediary metabolism (Duttagupta et al. 2005).

Coupling of intracellular signalling pathways and post-transcriptional control

Nitrogen starvation and nutrient sensing

The TOR pathway is distinctive for having global effects on the translational machinery. Amino acid starvation, or addition of the TOR inhibitor rapamycin rapidly shuts down synthesis of ribosomal components. The TOR pathway acts on the translational machinery at different levels (transcriptional, post-transcriptional and post-translational) and through multiple mechanisms, including regulation of RNA polymerases I and II, activation of eIF4E through hyperphosphorylation of 4E-BPs, and the ribosomal protein S6 kinase. In yeast, two eIF4E-binding proteins have been identified, Caf20 and Eap1. As eap1 mutants are rapamycin-resistant, this opens the possibility that the modulation of translational activity by the TOR pathway are in part channelled through phosphorylation of Eap1, in analogy with 4E-BP control in mammalian cells (Cosentino et al. 2000). Signalling through the TOR pathway also promotes translation through the phosphatase Sit4, which promotes conversion of eIF2α to its unphosphorylated, active, form (Cherkasova and Hinnebusch 2003; Rohde et al. 2004). Beside these general effects on translation of mRNA, the TOR pathway exerts post-transcriptional control on specific mRNAs. In mammalian cells, translation of mRNAs containing a pyrimidine-rich stretch in the 5′-UTR (Table 1), is promoted by TOR activity (Jefferies et al. 1994). This sequence feature is present in many mRNAs encoding proteins needed for translation, including ribosomal proteins, again linking the TOR pathway to protein synthesis in general.

Table 1 Some elements in mRNA sequences important for regulation of stability and translation
Full size table

In budding yeast, passage from G1 to S-phase is critically dependent on the protein level of the cyclin Cln3. Translation of CLN3 mRNA is reduced under conditions of low protein synthesis rate and slow growth, and this regulation is dependent on a uORF in the 5′-UTR (Polymenis and Schmidt 1997). The G1 arrest induced by rapamycin can be overcome by replacing the 5′-UTR of CLN3 with a heterologous sequence lacking this uORF (Barbet et al. 1996). The cell-cycle gene CDC33 encodes the cap-binding protein eIF4E. cdc33-1 mutants arrest in G1 with low Cln3 levels. Substitution of the 5′-UTR with the same heterologous sequence as above in the CLN3 mRNA abolishes the G1-specific arrest in cdc33-1 mutants (Danaie et al. 1999), implying that translation of CLN3 is the limiting factor for cell cycle progression in such mutants, and that eliminating the uORF in CLN3 enhances translation sufficiently to compensate for the eIF4E deficiency of cdc33-1 cells.

A case that is parallel in many aspects to budding yeast Cln3 has been found in fission yeast. There, the mRNAs encoding two key cell cycle regulators, the Cdk1 phosphatase Cdc25 and the mitotic cyclin Cdc13, both have extensive stem-loop structures in their 5′-UTRs (Daga and Jimenez 1999). The translation of both the cdc25 + and cdc13 + mRNAs are highly sensitive to the expression levels of tif1 +, which interestingly encodes eIF4A, an RNA helicase and a component of eIF4F which targets the 5′ end of mRNAs and unwinds secondary structures in the 5′-UTR.

One study investigated the global effects in yeast on translation by fusel alcohols (e.g. butanol), which accumulate under nitrogen starvation as a result of amino acid breakdown (Smirnova et al. 2005). One might then expect that amino acid starvation and butanol stress would cause similar post-transcriptional effects. Instead, they cause persistent translation, refractory to the general translational down-regulation, of distinct and mostly non-overlapping sets of mRNAs. Intriguingly, eIF2B is targeted in both cases, which underscores that additional mechanisms for control of translation under these stress conditions are likely to exist.

Glucose leads to destabilisation of mRNAs required for gluconeogenesis (the critical enzymes fructose-1,6-bisphosphatase, encoded by FBP1, and phosphoenolpyruvate carboxykinase, encoded by PCK1, are specific for gluconeogenesis and are specifically targeted in this way). This is dependent on the low glucose sensor Snf3 and the phosphatase Reg1 (Yin et al. 2000). Interestingly, different pathways seem to regulate this effect under different glucose concentrations. In conditions of high glucose, the high glucose sensor Rgt2 and the Ras/cAMP pathway are required, whereas for signalling in low glucose concentrations, the low glucose sensor Snf3 is needed, but the Ras pathway is dispensable.

Stress-activated MAPK pathways

Eukaryotic cells possess multiple three-layered MAP kinase cascades, which act in parallel in response to different stimuli. The stress-activated MAP kinases (SAPK’s) are stimulated by a variety of conditions, including oxidative and hyperosmotic stress, UV irradiation, and temperature changes. In mammalian cells, there are two types of SAPKs: the c-Jun NH2-terminal protein kinase (JNK) and the p38 subfamilies. Plants possess a large number of MAP kinases, but their possible role in post-transcriptional regulation has not been established. The SAPK in S. cerevisiae is called Hog1, and its homologue in Sz. pombe is Sty1. A MAP kinase can bind to and activate a downstream kinase (MAPK-activated protein kinase; MAPKAPK). Mammalian MAPKAPK’s have been directly implicated in several instances of post-transcriptional control. In yeast, there is a MAPK-activated protein kinase, which shares some homology to mammalian MAPKAPK’s. The primordial gene encoding this kinase has been independently duplicated in S. cerevisiae and Sz. pombe (Hughes and Friedman 2003) to produce the paralogues Rck1 and Rck2 in the budding yeast (Dahlkvist and Sunnerhagen 1994) and Srk1/Mkp1 and Cmk2/Mkp2 in the fission yeast (Alemany et al. 2002; Asp and Sunnerhagen 2003; Smith et al. 2002).

Regulation through A/U-rich elements

The best characterised sequence elements in mRNAs involved in regulation of stability and translation are the A/U-rich elements (AREs), first defined in mammalian cells (for review, see Bevilacqua et al. 2003; Zhang et al. 2002). These elements (see Table 1), most often found in the 3′-UTR, recruit proteins with an influence on mRNA translation and stability. These proteins in turn can bind to components of the exosome to initiate 3′–5′ degradation of selected mRNAs (Anderson and Parker 1998; Chen et al. 2001).

In yeast, translation of the MFA2 and the TIF51A mRNAs is down-regulated in glucose medium (Vasudevan et al. 2005). By contrast, the stability of TIF51A mRNA is increased in glucose whereas MFA2 mRNA is unaffected by carbon source (Vasudevan and Peltz 2001). Both the effects on mRNA stability and translation rate are dependent on an intact HOG pathway and on ARE’s in the 3′-UTRs of the transcripts. In mammals, expression of several interleukins is under control at the post-transcriptional level of SAPK pathways. Interleukin-2 (IL-2) mRNA is stabilised after activation of the JNK pathway, and this stabilisation is dependent on a JNK-responsive element (JRE) in the 5′-UTR of the IL-2 mRNA (Chen et al. 1998, 2000). The mRNAs for other interleukins, IL-6 and IL-8, become stabilised by cytokine signalling mediated by another SAPK, p38. In this case, also expression of the kinase downstream of p38, MAPKAPK-2 or MK2, is able to mediate this effect, whereas a mutated, kinase-dead, version of MK2 is unable to do so (Winzen et al. 1999). This role of SAPK’s is not restricted to cytokines, as inhibition of the p38 pathway destabilises cyclooxygenase-2 mRNA in several different cell types (Lasa et al. 2001). Evidence for regulation of mRNA stability involving ARE’s through the related SAPK JNK also exists. However, although there are numerous cases where similar effects are mediated from the SAPK p38 through the downstream kinase MAPK-activated protein kinase 2 (MAPKAPK-2 or MK2) (Kotlyarov et al. 1999; Lasa et al. 2000; Neininger et al. 2002; Winzen et al. 1999), this is not the case for JNK. An MK2-dependent phosphorylation of the ARE-binding protein tristetraprolin releases it from stress granules and inhibits degradation of mRNAs containing ARE’s (Stoecklin et al. 2004).

In mammalian cells exposed to certain but not all environmental stresses or mitogens, the MAPKAP kinase Mnk1 phosphorylates the cap-binding protein eIF4E. Mnk1 in turn can be activated by the p38 or ERK MAP kinases (Banko et al. 2004; Wang et al. 1998; Waskiewicz et al. 1999). SAPK-directed modulation of mRNA stability can occur also under a normal developmental process. In myoblasts, activated p38 phosphorylates and inactivates the mRNA-destabilising protein KSRP, leading to stabilisation of mRNAs critical for differentiation to myocytes (Briata et al. 2005).

It is not far-fetched to think of coordinated post-transcriptional control of functional groups of genes. In humans, expression is regulated by iron levels through iron response elements (IRE’s) in the UTR’s of mRNAs encoding proteins in iron metabolism. During iron starvation, the RNA-binding protein IRP1 becomes activated and attaches to the IRE in the 5′-UTR of selected mRNAs, preventing translation of proteins such as iron exporters and storage proteins. At the same time, IRP1 binds to the 3′-UTR and stabilises the mRNA for proteins required at iron scarcity such as transferrin receptor, facilitating uptake of iron from the exterior (Hentze et al. 2004). It has recently been established that a large class of genes involved in iron metabolism is under coordinated post-transcriptional control by the Cth2 protein in yeast (Puig et al. 2005). Cth2 displays sequence homology to the human ARE-binding protein tristetraproline (TTP) in the TZF region. It was also shown that a majority of the mRNAs that are dependent on Cth2 for post-transcriptional down-regulation under iron limitation carry ARE elements in their 3′-UTR. Thus, the IREs in this organism appear to be identical to, or a subset of, AREs (see Table 1). Although it is clear that a coordinated control of mRNA stability occurs in response to changes in iron levels in yeast, the responsible signalling pathway(s) have not been identified.

Other instances of MAPK-mediated post-transcriptional control

Oxidative stress in fission yeast causes phosphorylation of eIF2α, with concomitant reduction of protein synthesis. This stress-induced phosphorylation has been reported to be substantially increased in sty1 mutants, suggesting that the SAPK pathway is involved in modulating the global translational response to stress (Dunand-Sauthier et al. 2005).

In budding yeast, an Rck2-dependent phosphorylation of EF-2 upon hyperosmotic shock has been demonstrated to occur, which in turn is dependent on phosphorylation of Rck2 by the MAPK Hog1 (Teige et al. 2001). In hog1 and rck2 mutants, the global adaptation (initial down-regulation and later recovery) of translation upon hyperosmotic and oxidative stress is perturbed (Swaminathan et al. 2006; Teige et al. 2001; Uesono and Toh 2002). In rck2 mutants, a large number of genes encoding ribosomal proteins and proteins required for ribosomal assembly and modification are deregulated both on the transcriptional level and with respect to polysomal association (Swaminathan et al. 2006).

The Sz. pombe RNA-binding protein Csx1 is phosphorylated upon oxidative stress in a Sty1-dependent way (Rodriguez-Gabriel et al. 2003). Interestingly, this phosphorylation is specific for the type of external stress, and does not occur in, e.g. hyperosmotic shock. This leads to stabilisation of mRNA encoding Atf1, a transcription factor responsible for stimulation of transcription a large number of stress-induced genes. Two other RNA-binding proteins, Cip1 and Cip2, have subsequently been identified which act in opposition to Csx1 (Martin et al. 2006); the sensitivity of a csx1 deletion to oxidative stress is rescued by a cip1 or cip2 deletion. There is as yet no evidence for Sty1-dependent phosphorylation of Cip1 and Cip2, however. As atf1 + mRNA is the only identified direct target of Csx1, Cip1, and Cip2, it is not clear if the mechanism involves ARE’s.

So far, we have considered regulation at the post-transcriptional level as one of the outputs of signalling chains in the cell, including MAP kinase cascades. There is one interesting example of post-transcriptional regulation acting within this signalling pathway itself. In fission yeast, the RNA-binding protein Rnc1 participates in a negative feedback mechanism leading to down-regulation of the Pmk1 MAPK pathway after activation. This occurs through phosphorylation of Rnc1 by active Pmk1, inducing the binding of phosphorylated Rnc1 to pmp1 + mRNA, encoding a phosphatase which deactivates Pmk1. This leads to mRNA stabilisation with ensuing dephosphorylation and deactivation of Pmk1 (Sugiura et al. 2003).

DNA damage checkpoints

Proteins in the DNA-dependent cell cycle checkpoint pathway were primarily recognised for their post-translational action on cell cycle proteins, and for their crucial role in the transcriptional induction program following DNA damage and stalled replication. This pathway may also act at the post-transcriptional level, however. Signalling from replicatively damaged DNA through the checkpoint protein Dun1 destabilises RAD5 mRNA through interaction with the polyA nuclease (PAN) complex (Hammet et al. 2002). This leads to decreased Rad5 levels upon checkpoint activation, favouring error-free DNA repair. This action is mediated through the FHA domain of Dun1 in its interaction with the Pan3 subunit of the polyA nuclease complex. The up-regulation of UV damage endonuclease (UVDE) activity in Sz. pombe rad9 checkpoint mutants has been ascribed to post-transcriptional effects, on mRNA stability, translation, or both (Davey et al. 1998).

There are also examples of post-transcriptional effects following DNA damage where the signalling pathway remains unidentified. UV irradiation of mammalian cells causes down-regulation of translation through Gcn2-dependent phosphorylation of eIF2α (Deng et al. 2002; Jiang and Wek 2005). Even though UV radiation activates the JNK and p38 stress-activated MAP kinases, eIF2α phosphorylation does not require either of these pathways. Conversely, UV-induced activation of JNK or p38 does not require Gcn2 (Deng et al. 2002), and so there is presently no evidence for a link between these phenomena. In yeast, however, treatment with the DNA damaging agent methyl methane sulfonate (MMS) increases Gcn4 translation in a Gcn2-dependent way (Natarajan et al. 2001). It should be emphasised that it has not been shown that DNA damage as such is the initiating event for Gcn2 activation in these experiments.

Protein kinase C

A direct coupling of protein kinase C (PKC) signalling and translation in mammalian cells is strongly suggested by the physical association of both the PKC receptor protein RACK and PKCβII to eIF6, a protein implicated in dissociation of the 40S and 60S subunits. Stimulation of PKC signalling promotes joining of the ribosomal subunits (Ceci et al. 2003). The significance of RACK/PKC signalling is perceived to lie in cell-cell adhesion and localised translation at points of cellular growth.

Global studies of translational efficiency

Global investigation of cognate protein and mRNA levels in S. cerevisiae have pointed out that the degree of correlation is so low that major post-transcriptional regulation has to be invoked (Futcher et al. 1999; Ghaemmaghami et al. 2003; Gygi et al. 1999; Washburn et al. 2003). This discrepancy is most pronounced for weakly expressed genes (Pradet-Balade et al. 2001). Different cellular compartments have been reported to have different characteristics with respect to translational regulation (Beyer et al. 2004). An investigation of pre-existing datasets describing mRNA abundance, protein abundance and translational efficiencies indicate that there is a positive correlation between mRNA abundance and ribosome density. Thus, mRNAs encoding proteins in energy-yielding processes tend to be translated more efficiently; those where the levels of the protein product varies rapidly (e.g. ribosomal proteins) are of relatively high abundance but their translation can be strongly up-regulated when needed (“translation on demand”; Beyer et al. 2004; Washburn et al. 2003).

Several array studies indicate that changes in the translational state and stability may affect a quite broad range of mRNAs. Blocking of the Ras and Akt pathways in mammalian brain cells leads to rapid regulation of polysomal association of selected mRNAs, arguably more so than seen in the abundance in total mRNA. A significantly higher number of transcripts relevant for Ras and Akt signalling were found to be altered in polysomal mRNA than in the total mRNA fraction (Rajasekhar et al. 2003). An early study identified 1% of transcripts in fibroblasts as being translationally regulated upon mitogenic stimulation (Zong et al. 1999). During transition from an epithelial to a fibroblastoid phenotype, 15% of all induced or repressed transcripts were exclusively translationally regulated, while for 7%, both a transcriptional and a translational component was found (Jechlinger et al. 2003). A recent study of changes on the mRNA level after exposure of tumour cells to ionising radiation found as much as 90% of all induced and 80% of all repressed mRNAs to be regulated primarily on the translational level (Lü et al. 2006). In yeast, equal numbers of mRNAs were found to be regulated on the translational versus the transcriptional level upon amino acid starvation and butanol stress (Smirnova et al. 2005).

Comparing steady-state levels of mRNA species in the total RNA pool and their levels in polysomal RNA under one particular set of conditions gives a static view on translational control. There is a positive correlation between high expression level of an mRNA and its tendency to be highly translated (Beyer et al. 2004). Several genome-wide studies of mRNA translation profiles have been performed in yeast, aiming to identify global changes that occur upon changes in growth conditions or external stress, e.g. (Arava et al. 2003; Kuhn et al. 2001; Preiss et al. 2003; Smirnova et al. 2005; Swaminathan et al. 2006). This gives a dynamic perspective on translational regulation. The strongest overall trend is to have transcriptional and translational control reinforce each other for a particular gene, a phenomenon dubbed “homodirectional changes” (Preiss et al. 2003). For a small number of genes, there are instead “counterdirectional” changes. The budding yeast genes CPA1, GCN4, HAC1, and ICY2 are more efficiently translated under conditions of stress (Beilharz and Preiss 2004; Hinnebusch and Natarajan 2002; Kuhn et al. 2001; Messenguy et al. 1983). For GCN4, this control is exerted through phosphorylation of eIF2α and different translation of uORFs as already described. In mammalian cells, translation of the ATF4 mRNA (also encoding a transcription factor) is likewise positively controlled by eIF2α phosphorylation and also involves translation of uORFs. Interestingly, the mechanism is similar to that of yeast GCN4 in that it involves a more 5′ uORF which directs scanning and downstream reinitiation, and a second, inhibitory, uORF closer to the main ORF (Vattem and Wek 2004). Another mechanism for translational control is through phosphorylation of eIF4E binding proteins, leading to release of active eIF4E. An active TOR pathway and many mitogens increase this phosphorylation. This regulation can have gene-specific effects, as translation of the PGC1 mRNA is selectively increased in mice lacking eIF4E binding protein 1 (Tsukiyama-Kohara et al. 2001).

The global studies indicate that a considerable fraction of mRNA species change their translational efficiency in response to stress, hormones or nutrient availability. It is thus likely that beside these few well-characterised cases, other as yet distinguished mechanisms regulate potentially large classes of mRNAs on the translational level. This regulation may be surprisingly specific, as even seemingly related stress conditions will enhance polysomal association of different mRNA subsets (Smirnova et al. 2005). Further, their data suggest co-regulation of translation and transcription factors by the same signalling pathways.

Concluding remarks

The extent of posttranscriptional regulation could be different between yeast and more complex organisms. It is well recognised that the scarcity of introns provides fewer opportunities for alternative splicing in yeast. Of note is also the fact that although mammalian mRNAs encoding ribosomal proteins contain polypyrimidine tracts in their 5′-UTRs which dictate their translational regulation (Table 1), the corresponding yeast mRNAs lack these sequences and are instead regulated on the transcriptional level (Warner 1999). RNAi presents a wealth of previously unrecognised possibilities for short-term and long-term regulation of expression. This mechanism is almost ubiquitous in eukaryotes with the notable exception of hemiascomycetes (budding yeasts), among which the investigated species invariably have lost the genes encoding proteins required for the RNA processing steps in the RNAi pathway and several genes required for heterochromatin formation (Aravind et al. 2000; Axelsson-Fisk and Sunnerhagen 2006; Cerutti and Casas-Mollano 2006). Based on the distribution among eukaryotic lineages of components of the RNAi machinery, it has been suggested that its primordial functions included targeted transcript degradation, and that other functions such as guided DNA methylation could be more recent specialisations (Cerutti and Casas-Mollano 2006). There are recent hints of previously unsuspected links between the RNAi machinery and regulation of mRNA turnover. Thus, microRNA regulation of mRNA stability can be mediated through ARE elements in mammalian cells (Jing et al. 2005). It has also been found that mammalian P bodies contain protein and RNA elements of the RNAi machinery (Liu et al. 2005; Sen and Blau 2005). Knowing that formation of stress granules is downstream of phosphorylation of eIF2α, which is also a focal point of intracellular signalling through TOR and other pathways, it appears likely that temporary storage of mRNAs in response to stress is controlled through one or more of these intracellular signalling pathways. Great efforts are currently made to define the molecular mechanisms of such interactions between pathways.

So far, only few verified examples of uORF control of translation exist, and the extent of this control mechanism is unknown. The number of conserved uORFs emerging from comparative genomics studies suggests that uORF control could be quite widespread. Monitoring movement of mRNAs of interest between the polysomal fractions and those containing free mRNAs across several array experiments, representing many different types of environmental and hormonal inputs, should be an efficient experimental way to identify mRNAs regulated by uORFs.

A global investigation of protein half-lives in budding yeast indicates that the protein products of genes that are co-regulated on the transcriptional level tend to be co-regulated also on the level of protein turnover (Belle et al. 2006). It would be interesting to see if a similar, or even stronger, correlation could be found between post-translational and post-transcriptional regulation. In more general terms, a fruitful approach to identification of mRNA classes subject to post-transcriptional regulation should be to measure translational activity (as polysomal association) and mRNA stability (after Pol II arrest) in the same experiment. Cross-species comparisons will add to the discrimination power of this approach, and may lead to the identification of sequence determinants, conserved in evolution, that are responsible for mRNA stability and translational efficiency.