Cell, Vol. 62, 15-24,
July 13, 1990, Copyright
0 1990 by Cell Press
Do the Poly(A) Tail and 3’ Untranslated Region Control mRNA Translation? Richard J. Jackson and Nancy Standart Department of Biochemistry University of Cambridge Cambridge CB2 1QW England
From the date of its discovery in 1970, the 3’ poly(A) tail of eukaryotic mRNA has taken on the character of a structure in search of a function. Candidate functions favored in those far-off days were: mRNA transport from nucleus to cytoplasm, the control of mRNA stability, the compartmentalization of mRNA in the cytoplasm and the question of its association with the cytoskeleton, and, as a more remote possibility, a role in the actual mechanism or control of translation (reviewed by Brawerman, 1981). Over the intervening years, faith in the importance of the poly(A) tail as a determinant of mRNA stability has been maintained and recently rewarded by interesting and provocative results (reviewed by Bernstein and Ross, 1989). In the meantime, support for a role of the poly(A) tails and the 3’ untranslated region (UTR) in the control of translation has been more fickle, but here too there has been a recent upsurge of interest and a wave of significant new results. These new data are drawn from a wide variety of organisms and biological phenomena; this review aims to gather these different strands of evidence together, while taking care not to make unjustified extrapolations from one organism or system to another. Poly(A) Tail Synthesis and Metabolism Although the details are outside the scope of this review, a brief account of the polyadenylation process in the nucleus is probably useful. In mammalian cells the primary hnRNA transcript is cleaved at a point specified by its location within 30 nucleotides downstream of an AAUAAA motif and just upstream of a GU-rich segment that shows less sequence conservation (reviewed by Humphrey and Proudfoot, 1988). This cleavage is rapidly followed by the addition of A residues to the free 3’ end in what appears to be a two stage process. The initial stage shows a strong requirement for the AAUAAA motif, but once some ten A residues have been added, further polyadenylation seems dependent only upon the preexisting oligo(A) tail (Sheets and Wickens, 1989). Pulse-labeling experiments show that the newly synthesized mRNA detected in the nucleus carries poly(A) tails with a narrow size distribution at about 250 nucleotides in mammalian cells but nearer 100 nucleotides in lower eukaryotes. It is not clear what limits the length of the poly(A) tail, but it is unlikely to be due to an intrinsic property of the poly(A) polymerase (Sheets and Wickens, 1989). On transport to the cytoplasm, the poly(A) tail seems to undergo a slight shortening but still shows a narrow size distribution (Brawerman, 1981). Thereafter, the length becomes shorter and more heterogeneous, probably because the rate of shortening differs for different mRNA
Review
species (see below). Some quite dramatic cases of poly(A) lengthening have been observed in developmental situations, as will be discussed later, but even in somatic cells there is some end addition of A residues in the cytoplasm that partly counteracts the shortening process (Brawerman, 1981). In the presence of actinomycin to inhibit further mRNA synthesis, labeling of cytoplasmic mRNA poly(A) reaches steady-state within about 10 min, representing an average of about five A residues added to each tail. Poly(A) Binding Protein (PABP) The question of any role of the poly(A) tail in controlling mRNA stability or translation cannot be considered in isolation from the function of the poly(A) binding protein (PABP), a -70 kd protein common to most if not all eukaryotes. Cloning of the PABP genes of yeast, humans, and Xenopus has revealed a common domain structure in the N-terminal two-thirds of the protein, which consists of four repeat units, each about 90 amino acid residues in length and highly conserved in all three organisms (Adam et al., 1986; Sachs et al., 1986; Grange et al., 1987; Zelus et al., 1989). The sequence of the C-terminal one-third is conserved in humans and Xenopus, but not in yeast. In yeast, it has been possible to demonstrate that the PABP gene is essential for viability (although certain suppressor mutations can obviate the requirement). However, the complete gene can be replaced by truncated forms consisting of just a single domain with little effect on cell growth at 30% but a reduction of cell growth at 37°C (Sachs et al., 1987). Digestion of poly(A)-PABP complexes with nonspecific nucleases results in fragments with PABP bound to a length of 27 A residues (Baer and Kornberg, 1983). However, the common assumption that the minimal binding site for PABP is 27 residues may well be incorrect, as studies with poly(A) of different lengths showed that a 12 nucleotide fragment was required for efficient binding, and little increase in binding was seen with longer lengths, irrespective of whether the full-length PABP or a singledomain derivative was studied (Sachs et al., 1987). The 27 nucleotide repeat relates to the actual spacing of the PABP molecules on the poly(A) tail and not to the minimum size of the binding site. With the single domain derivative, this spacing decreases to about 13 nucleotides. Binding of PABP to poly(A) is very stable even in high salt concentrations, but this tight binding should not be regarded as immobilizing the PABP on the poly(A). There is considerable evidence that PABP molecules can migrate or transfer between poly(A) tracts, a migration that appears
to require
the
multidomain
protein
structure
and
is
not seen with the single-domain derivative (Sachs et al., 1987). It seems likely that the poly(A) tails of all cytoplasmic mRNAs in all eukaryotes are associated with PABP, but Xenopus oocytes may be an exception. lmmunoblotting methods failed to detect PABP in oocytes or early embryos, and although more sensitive methods showed that it is actually present in oocytes, the question remains
Cdl 16
open as to whether there is sufficient PABP to cover all of the mRNA poly(A) tails (Zelus et al., 1989). Maternal PABP mRNA appears to be destroyed at oocyte maturation, and the mRNA only reappears at the neurula stage of embryogenesis, the stage when PABP itself becomes easily detectable and when there is a large increase in poly(A)+ mRNA. The Role of Poly(A) Tails, 3’ UTR Sequences, and PABP in Controlling mRNA Stability A brief consideration of mRNA degradation mechanisms is necessary at this stage, since any claims for an influence of poly(A) tails and 3’ UTR sequences on translation efficiency always provoke questions as to whether the effects can be explained in terms of mRNA stability. The idea that poly(A) tail shortening and its ultimate removal is the prelude to mRNA degradation originates from the fact that the length of the tails decreases as mRNA ages in the cytoplasm, and in the case of shortlived mRNAs such as c-fos, poly(A) tail shortening is exceptionally rapid (Wilson and Treisman, 1988). It is also generally agreed that apart from species with shortened poly(A) tails, it is hard to detect any other mRNA degradation intermediates. The few exceptions are short-lived histone H4 mRNA and c-myc mRNA intermediates missing small segments of extreme 3’ UTR sequences (Ross and Kobs, 1988; Ross et al., 1986; Brewer and Ross, 1988). This suggests that once the poly(A) tail has been removed, degradation of the body of the mRNA normally proceeds rapidly and processively in the general 3’+5 direction, catalyzed either by an exonuclease or by an endonuclease that removes small fragments. For a few mRNA species, notably j3-actin mRNA, poly(A) shortening is not the prelude to rapid degradation, and there is a substantial pool of actin mRNA with, at most, a very short oligo(A) tail that is not destined for rapid destruction (Brawerman, 1981). It seems possible that these mRNAs have structures near the 3’end that make them particularly resistant to the degradative processes. Studies of mRNA degradation in a cell-free system that seems to reproduce faithfully in vivo degradation mechanisms show that the stabilizing effect of the poly(A) tail requires PABP in a way which suggests that the poly(A) tail is degraded only when it is naked and unassociated with PABP (Bernstein and Ross, 1989; Bernstein et al., 1989). This implies that the poly(A) tails of short-lived mRNAs such as c-fos and c-myc, which undergo very rapid shortening, must be naked for a higher proportion of the time than stable mRNAs such as globin. A major determinant of the rapid turnover of these mRNAs as well as lymphokine mRNAs is the multiple copies of AU-rich motifs (consensus sequence UAUUUAU) common to the 3 UTRs of these mRNAs (Caput et al., 1986; Wilson and Treisman, 1988; Jones and Cole, 1987). Transfer of these segments to the 3’ UTR of globin mRNA dramatically reduces its stability (Shaw and Kamen, 1986). It has been suggested that PABP readily migrates from the poly(A) tract to these AU-rich segments of the 3’ UTR, leaving the poly(A) tract naked and vulnerable to degradation for a high proportion of the time (Bernstein and Ross, 1989). An
alternative model postulates base pairing between the poly(A) tail and the UAUUUAU motifs in the 3’ UTR, which might exclude PABP from binding to the poly(A) tail and allow nuclease attack at the points of mismatch (Wilson and Treisman, 1988). A relationship between 3’ UTR structures and degradation is also demonstrated by the stabilization of transferrin receptor mRNA under conditions of iron deficiency, which is thought to be due to the binding of a regulatory protein to five repeated sequence motifs in the 3’ UTR (Casey et al., 1988); it is not yet clear whether this binding affects the rate of poly(A) shortening or the degradation of the body of the mRNA that follows poly(A) tail removal. However, events and sequences in other parts of the mRNA can also affect degradation rates: the selective destruction of tubulin mRNA promoted by unpolymerized tubulin subunits (Yen et al., 1988); the destabilization of c-fos mRNA promoted by a sequence element within the coding segment that acts independently of the 3’ UTR AU-rich motifs (Shyu et al., 1989); the influence of 5’-proximal sequences on c-myc degradation (Jones and Cole, 1987); the alterations in mRNA stability when the translational stop codon is moved to a different site in the mRNA (Marzluff and Pandey, 1988; Daar and Maquat, 1988); and the stabilization of some mRNAs such as c-fos and histone mRNAs that occurs when protein synthesis is inhibited by cycloheximide (Marzluff and Pandey, 1988; Wilson and Treisman, 1988). If all of these determinants influence mRNA degradation rates by affecting the rate of poly(A) shortening and/or the rate of subsequent degradation from the 3’ end of the body of the mRNA, this implies some sort of “communication” or functional interaction between the 3 end of the mRNA and far upstream segments, an interaction that could also be significant in the control of translation. The discussion above has focused on evidence from mammalian systems, and there are some doubts as to whether it can be extrapolated in its entirety to lower eukaryotes such as yeast. The short-lived mRNAs in yeast do not seem to share common 3’ UTR sequence elements such as the AU-rich motifs of mammalian lymphokine and proto-oncogene mRNAs, and for one rapidly degraded species, the MATal mRNA, the primary determinant of the instability is a short segment in the coding sequence with a high frequency of rare codons (Parker and Jacobson, 1990; Herrick et al., 1990). Moreover, there is some evidence that PABP in yeast, far from preventing poly(A) shortening, may actually promote it (Sachs and Davis, 1989). Effect of Poly(A) Tails and 3’ UTR Sequences on mRNA Stability and Ranslation in Xenopus Oocytes It is also not clear to what extent the mechanism of mammalian cell mRNA degradation is operative in Xenopus oocytes. This is an important issue since experiments involving the microinjection of poly(A)+ and deadenylated mRNAs into oocytes have provided much of the evidence on the question of a role of poly(A) tails in translation. Whereas the earlier studies used polynucleotide phosphorylase to deadenylate poly(A)+ mRNAs and poly(A)
Review: 17
Poly(A)
Tails
and 3’ UTR Sequences
polymerase for (re)addition of tails, more recent work has used in vitro transcripts of cloned genes, which has the advantage that poly(A) tails of defined length can be added, although in many such experiments the cloning strategies have given tails that are not purely poly(A). In interpreting such results it is important to distinguish between assays of functional half-life (determined by following the rate of synthesis or accumulation of the translation product) and chemical half-life (determined by following the fate of the input RNA). If polyadenylation influences both translation efficiency and mRNA stability, the functional half-life cannot be assumed to be equivalent to the chemical or structural half-life. The most influential result in this area, but unfortunately the least typical, has been the original report that the functional half-life of deadenylated globin mRNA was much shorter than that of poly(A)+ globin mRNA (Huez et al., 1974). Readenylation experiments showed that a tail of 32 A residues was sufficient to confer a long functional halflife, while addition of only 16 residues gave the same results as fully deadenylated mRNA (Nude1 et al., 1976). In addition, polyadenylation of HeLa cell poly(A)- mRNA prior to microinjection greatly increased the functional lifetime of the histone mRNAs, which was only a few hours in the absence of a poly(A) tail (Huez et al., 1978). In contrast, direct assay of the fate of the microinjected RNAs showed a much smaller difference between the stability of deadenylated globin mRNA and globin mRNA carrying either a long poly(A) tail or a 3’ tail consisting of A&& (Drummond et al., 1985; Galili et al., 1988). In similar assays with other mRNA species, the effect of polyadenylation on chemical stability over the first 24 hr following microinjection was very small, and it was only by 48 hr that significant differences were evident (Drummond et al., 1985; Galili et al., 1988). However, polyadenylation did have significant effects on translation efficiency in these experiments. The addition of long poly(A) tails increased the rate of lysozyme mRNA translation in oocytes by 20fold and chymosin mRNA translation by lO-fold (Drummond et al., 1985). A 3’ tail of A55C20 on zein mRNA stimulated translation by over 20-fold and a similar tail on j3-globin mRNA gave a 6-fold stimulation (Galili et al., 1988). All of these assays were performed at 24 hr after microinjection and took into account differences in mRNA stability, with the slight reservation that as stability is assessed by the size and amount of surviving input RNA, subtle changes in RNA structure cannot be ruled out. In experiments where the time course of protein synthesis has been examined, as in the original work with globin mRNA (Huez et al., 1974) and more recent studies using zein mRNA (Galili et al., 1988) a common feature is that the deadenylated mRNA is translated with at least 50% of the efficiency of the poly(A)+ form during the first few hours after injection. Subsequently, the efficiency of translation of the poly(A)+ mRNA seems to increase, since the mRNA moves into larger polysomes (Galili et al., 1988). In contrast, the deadenylated mRNA remains in small polysomes and its translational efficiency tends to decrease. These results go some way toward explaining the discrepancies between the functional and chemical stability
of microinjected poly(A)- mRNA. They have been interpreted in terms of a model that distinguishes between translation initiation, which is envisaged as random initiation by ribosomal subunits present in the free subunit pool, and reinitiation, where a ribosome that has just completed translation of the open reading frame has a high probability of (re)initiating translation of the same mRNA molecule. The poly(A) tail is postulated to promote the reinitiation process (as opposed to the random selection pathway), presumably through promoting some sort of interaction between 3’- and Y-proximal elements of the mRNA, such as to increase the probability that ribosomes recycle on the same mRNA (Galili et al., 1988). To what extent are these results typical of other eukaryotic cells? Electroporation has been used to introduce into plant protoplasts j3-glucuronidase (GUS) mRNA derivatives with different 5’and 3’ UTR sequences (Gallie et al., 1989). The presence of an AP5 tail increased the level of GUS activity detectable 8-24 hr after electroporation, the increase ranging from 12-fold to over 30-fold depending upon the plant species. Expanding the length of the tail to Az5GUUAz5UUUAAAGAAUU produced a further increase close to 2-fold in all species. Apparently a short length of non-poly(A) sequences downstream of the poly(A) tract was not detrimental, but longer segments of extraneous sequences largely negated the stimulatory effect of the poly(A) tract, especially in tobacco and maize. As for mRNA stability, the addition of a tail of 50 (mostly) A residues increased the half-life in tobacco protoplasts from about 30 min to 70 min, but this stabilization seems totally insufficient to account for the 30-fold increase in the yield of GUS activity. Besides the poly(A) tail itself, sequences in the 3’ UTR can also influence translation efficiency in vivo. The most striking case is that of the human p-interferon (IFN) mRNA, which is translated very inefficiently in Xenopus oocytes. Exchanging the UTR segments between different genes showed that this inhibitory effect lies mainly in the 3’UTR, with only a minor effect of the 5’ UTR. There is clearly some additional contribution from the coding sequence, since the IFN mRNA 3’ UTR transferred to the lysozyme coding sequence is significantly less inhibitory than when attached to the IFN coding sequence (Kruys et al., 1987). The negative elements in the 3’ UTR were shown to correspond to the AU-rich motifs common to lymphokine mRNAs which are thought to be determinants of the rapid turnover rates of these mRNAs in mammalian cells (Kruys et al., 1988, 1989). However, in Xenopus oocytes these sequences influence translation rather than degradation, for the microinjected IFN mRNA, whether polyadenylated or not, was stable for at least 24 hr. Moreover, the endogenous c-fos mRNA in oocytes has a half-life of many days, in contrast to its very rapid turnover in Xenopus fibroblasts (Mohun et al., 1989). It therefore seems that Xenopus oocytes may lack the mRNA degradation system found in mammalian somatic cells, or if they have such a system, then its activity would appear to be much lower and any coupling to sequences in the body of the mRNA must involve different sequence motifs.
Cdl 10
Thus, all of the experiments discussed in this section argue that in addition to any effect on mRNA stability, which seems generally small but quite variable between different mRNA species, polyadenylation can exert a direct influence on the translational efficiency of an mRNA in vivo. While this effect appears to be at the level of translation (re)initiation, it is in the nature of these assay systems that almost nothing can be ascertained as to the step in the initiation pathway that is stimulated. For evidence on this issue we must turn to other systems. The Influence of Poly(A) Tails, PABP, and 3’ UTR Sequences on Translation In Vitro Although the stimulatory effects of poly(A) tails on in vitro translation are much smaller than in intact Xenopus oocytes, such stimulation has been consistently observed in the reticulocyte lysate but not in the wheat germ system, a difference that may be related to the supposedly higher reinitiation efficiency in reticulocyte lysates. It is also of interest that the AU-rich sequence in the 3’ UTR of human p-interferon mRNA had no effect on translation in the wheat germ system but was inhibitory in the reticulocyte lysate, albeit to a lesser extent than in Xenopus oocytes (Kruys et al., 1987, 1988). In reticulocyte lysates, deadenylated ovalbumin mRNA was translated less efficiently (in competition against the endogenous globin mRNA) than the poly(A)+ form, and indirect assays showed that mRNA degradation was not the explanation for this difference (Doe1 and Carey, 1976). Deadenylation reduced elongation and initiation rates by about 20% and 40% respectively. In the nuclease-treated reticulocyte lysate, VSV mRNAs were translated at between 1.5fold and 3-fold higher efficiency if they were polyadenylated; however, neither the stability nor the ribosome transit times were affected, suggesting that differences in (re)initiation rates must be the primary cause (Jacobson and Favreau, 1983; Munroe and Jacobson, 1990). This was confirmed by studying the polysome distribution of differentially labeled poly(A)+ and poly(A)mRNAs added to the same assay: with either VSV mRNA or globin mRNA (in vitro transcripts with defined tails of pure poly(A)), the poly(A)+ form was preferentially incorporated into larger polysomes. With the globin mRNA derivatives, a tail of five A residues gave the same outcome as poly(A)- mRNA, 32 or 68 residues gave maximal stimulation, and 23 residues was intermediate (Munroe and Jacobson, 1990). When poly(A)+ and poly(A)- VSV mRNAs were compared in initiation complex formation assays, the yield of 40S-mRNA complexes (in the presence of edeine, which blocks subunit joining but also promotes aberrant scanning of 40s subunits) was the same with both species, but the yield of 80S-mRNA complexes (in the presence of anisomycin, which blocks peptide bond synthesis) with poly(A)- mRNA was only about half that obtained with the poly(A)+ form (Munroe and Jacobson, 1990). These results suggest that the absence of a poly(A) tail may reduce the rate of subunit joining, which converts a 40S-mRNA complex into an 80S-mRNA complex. However, a simple inhibition uniquely to this step should result in an accumu-
lation of 4OS-poly(A)mRNA intermediate complexes in the assays containing anisomycin. As this expectation was not fulfilled, we may need to seek a more complex interpretation. Another test for a role of poly(A) tails in translation is to determine whether free poly(A) inhibits cell-free protein synthesis. A complication here is that all homopolynucleotides inhibit at concentrations above about 30 kg/ml. However, in some circumstances, poly(A) but no other homopolymer (including poly(dA)) shows selective inhibition at much lower concentrations (5 ug/ml). This hypersensitivity was evident with poly(A) of average length of at least 51 nucleotides but not with poly(A) of 23 nucleotides. It has been seen in the nuclease-treated reticulocyte lysate or Lcell extract translating (deproteinized) poly(A)+ mRNAs, but was not seen with mRNPs (mRNA-protein complexes in which PABP is likely to be the main protein component) and nonpolyadenylated viral RNAs such as tobacco mosaic virus RNA, turnip yellow mosaic virus RNA, and reovirus mRNAs, or with endogenous mRNA in untreated cell extracts (Jacobson and Favreau, 1983; Lemay and Milward, 1986; Grossi de Sa et al., 1988; Munroe and Jacobson, 1990). Even with deproteinized globin mRNA, inhibition was only seen if the poly(A) was added at zero time rather than after a 5 min delay. It was therefore argued that poly(A) could inhibit only the translation of poly(A)+ mRNA that had not yet been assimilated into mRNA-protein complexes, presumed to be mRNA-PABP complexes. Consistent with this interpretation, the addition of PABP abolished the inhibition caused by poly(A) (Grossi de Sa et al., 1988). However, the quantity of PABP required to achieve this anti-inhibitory effect was very small, such that the model only seems self-consistent if PABP can be assumed to have a much higher affinity for the poly(A) tail of globin mRNA than for the free poly(A). When VSV poly(A)- mRNA or Dictyostelium poly(A)mRNA (mainly histone mRNA) was translated in the nuclease-treated reticulocyte lysate, or reovirus mRNA in L-cell extracts, low concentrations of poly(A) actually stimulated translation (Jacobson and Favreau, 1983; Lemay and Milward, 1986; Munroe and Jacobson, 1990). This suggests that poly(A)-PABP complex can act in trans to stimulate translation; such stimulation is presumably at the initiation step, but this remains to be proven. Taken as a whole, the in vitro evidence points to a possible role of the poly(A)-PABP complex in effecting a modest stimulation of initiation or reinitiation, which may be exerted at the subunit joining step. Control of this step of the initiation pathway has not been encountered before and was previously thought unfavorable for the “economy” of the translation machinery, since it would result in fruitless sequestration of 40s subunits and initiation factors. Genetic Evidence for a Role of PABP in Translation Initiation The role of PABP in yeast has been explored by placing the full-length PABP gene under the control of an inducible promoter or by replacing the full-length gene with a one-domain gene carrying a mutation that confers a temperature-sensitive phenotype, largely because it re-
Review: Poly(A) Tails and 3’ UTR Sequences 19
suits in reduced steady-state levels of the protein (Sachs and Davis, 1989). In both experimental systems, turning off the synthesis of PABP results in a gradual cessation of growth over about 10 hr and seemed to cause inhibition of translation initiation, although the polyribosome profiles actually show a reduction in polysome yield without the pronounced skew toward small polysomes typical of simple inhibition of initiation. As there may be avery small pool of residual PAPB, a possible explanation for these polysome profiles is that the few mRNAs still complexed with PABP are translated with normal efficiency, while the majority are translated inefficiently or not at all. Depletion of PABP also resulted in an imbalance in the levels of the accumulated free subunits, in that far more free 60s subunits than 40s were observed. The reason for this imbalance was not investigated, but if we can assume that the deficit in free 40s subunits is not due to some secondary lesion in small ribosomal subunit assembly, a possible explanation is that the “missing” 40s subunits were associated with mRNA in polyribosomes, as would occur if there were a defect in the subunit-joining step of the initiation pathway. Using the one-domain mutant, suppressors of the temperature-sensitive phenotype were isolated and further selected for cold sensitivity. These suppressors, which all allowed growth in strains with a complete deletion of the PABP gene, fell into seven complementation groups. One was shown to be a mutation in a protein of the 60s ribosomal subunit, and at least five of the other six groups can be assumed to affect the 60s ribosomal subunit since they conferred hypersensitivity to cycloheximide (Sachs and Davis, 1989). These results are provocative because they introduce a novel type of evidence to the problem and clearly show a relationship between the poly(A)-PABP complex and the role of the 60s subunit in translation. Further data delineating this relationship are eagerly awaited, but in the meantime possible explanations are that poly(A) tails uncomplexed with PABP inhibit protein synthesis initiation, perhaps at the subunit joining step, or that the poly(A) tail with bound PABP may be regarded as a positive effector of this step. Either explanation is consistent with the results obtained using the reticulocyte lysate system. The first would seem to be favored by the fact that many of the more stable mRNAs in yeast persist (and are presumably translated) for some time in the poly(A)- state (Herrick et al., 1990) but such an interpretation is probably premature as there is no information yet on whether the actual efficiency of translation of these mRNAs in vivo is affected by deadenylation. If the poly(A)-PABP complex does play a role in translation initiation in yeast, one might expect to be able to suppress PABP deficiency by mutations in the genes coding for initiation factors, but it is possible that the cold-sensitive selection step used by Sachs and Davis (1989) biased the selection in favor of suppressor mutations affecting ribosome structure and assembly. Another and surprising consequence of PABP depletion is that the length of the poly(A) tails increases from a heterogeneous size of 20-60 residues, typical of cytoplasmic mRNA in wild-type cells, to 90-100 residues, a length characteristic of nuclear mRNA (Sachs and Davis,
1989). Thus, PABP seems responsible for the shortening that normally occurs during mRNA transport from nucleus to cytoplasm. However, as the extragenic suppressors restored translation without causing shortening of the unusually long poly(A) tails, it seems unlikely that the role of PABP in the shortening process is directly linked to its role in translation. Correlations between the Polyadenylation State and the Efficiency of Translation of mRNAs during Development There are some striking examples where polyadenylation of a particular mRNA is tightly correlated with activation of its translation. Table 1 gives a representative list drawn from a wide range of organisms. Although most of these examples relate to developmental systems, there are sufficient exceptions to show that this is not a peculiarity of development. The changes in polyadenylation status are usually quite selective and far from nonspecific; in Xenopus oocyte maturation, for example, the majority of mRNAs undergo deadenylation and total cell poly(A) content declines by about 50% (Sagata et al., 1980), while a subset of mRNAs becomes polyadenylated. This polyadenylation seems to involve the addition of A residues to a preexisting short oligo(A) tail rather than de novo polyadenylation. In marine invertebrates, such as Spisula, the untranslated mRNAs in unfertilized eggs are capable of binding to poly(U)-Sepharose but not to oligo(dT)-cellulose, which implies a tail of some 10 A residues (Rosenthal and Ruderman, 1987). In higher eukaryotes the average size of the preexisting oligo(A) tails seems somewhat larger. The deadenylation of certain mRNAs, which occurs during development, generally correlates with a reduction in the rate of their translation in vivo (Table 1). There are, however, a few cases where this correlation seems to break down, as illustrated by the behavior of the Spisula 4Y4 mRNA species and of ribosomal protein mRNAs in developing Dictyostelium (Table 1). However, as there are certainly other translational control mechanisms independent of poly(A), the occurrence of some exceptions seems hardly surprising. There are also a few cases where translational activation actually correlates with deadenylation. Even during Xenopus oocyte maturation and early embryogenesis, where polyadenylation correlated with translational activation is the general rule, histone mRNAs become deadenylated at about the same time that their translation is stimulated. By the criterion of binding to oligo(dT)-cellulose, 50%-750/o of histone mRNAs are poly(A)+ in oocytes, but only about 25% are poly(A)+ in early embryos, when there is a very significant activation of translation of core histone mRNAs, followed somewhat later by activation of histone Hl synthesis (Ruderman et al., 1979; Woodland, 1980). At least in the case of histone H4 mRNA the maturation-dependent deadenylation involves complete removal of the rather short oligo(A) tails (Ballantine and Woodland, 1985). In mammalian spermiogenesis many mRNAs, including the protamine mRNAs, are stored as untranslated
oocyte
Mouse
rhythm
stimulation
preproinsulin
vasopressin 2-3
increase
x increase
unspecified
The top half lists cases reporting data on the distribution of the relevant mRNA between polysome-associated as changes in translation rate. a These mRNAs bind efficiently to poly(U)-Sepharose. implying tails of about 10 A residues. b These mRNAs bind poorly to poly(U)-Sepharose, implying tails of less than 10 A residues.
glucose
circadian
Rat insulinoma
Rat hypothalamus
increase increase
vasopressin oxytocin
osmotic
Rat hypothalamus
unspecified unspecified
little change? significant increase
30 40
1
H4 histones
x increase
x increase
6 x increase x decrease
>20
2-5
25 x increase
5 5
90 >90
10 50 0 15 90
0 0 0 75 50 0
Before
% of mRNA in polysomes
Translation
>20
Hsp90
2
phosphoribosyltransferase
plasminogen
hypoxanthine actin
tissue
1 protein
mRNAs
Ll
of Translation
Starfish (M. glacialis) oocyte maturation
Sea urchin (S. purpuratus) early embryogenesis
maturation
spermiogenesis
Mouse
protamine transition
most mRNAs ribosomal protein
switch from growth to development
Dictyostelium vegetative
protein
A9 84 D7 GIO ribosomal
Frog (X. laevis) oocyte maturation
reductase
and Activation
ribonucleotide cyclin A histone H3 a-tubulin 4YlO 4Y4
Polvadenvlation
mRNA
solidissima)
between
Clam (Spisula fertilization
I. Correlations
Organism
Table Data
100 25
100
100
100 100 100 100 5b
100 100 100 25 lob 10s
After
and nonpolysomal
60 50-75
5
0
40 80 50 100 100
108 0= 50a 100 100 100
Before
% of mRNA Bound to Oligo(dT)-Cellulose
Polyadenylation
states.
lower half shows
et al. (1987)
and Brandhorst
(1986)
and
examples
where
the data have
et al. (1986)
et al. (1988)
been given
Smith et al. (1988) Ruderman et al. (1979); Ballantine and Woodland (1985) Carrazana et al. (1986); Carter and Murphy (1989)
Standart
Bedard
Muschel
The
(1989)
Huarte et al. (1987); Vassalli et al. (1989) Paynton et al. (1988)
Kleene
Robinson
H4)
(1987)
(1988) Palatnik et al. (1984) Steel and Jacobson (1988)
McGrew et al. (1989) Hyman and Wormington
240
30
400 increase
0 (histone
and Ruderman
Dworkin et al. (1985); Dworkin Dworkin-Rash (1985)
Rosenthal
References
increases by 130 nucleotides
250 unspecified
up to 14
increase
July 13, 1990, Copyright
0 1990 by Cell Press
Do the Poly(A) Tail and 3’ Untranslated Region Control mRNA Translation? Richard J. Jackson and Nancy Standart Department of Biochemistry University of Cambridge Cambridge CB2 1QW England
From the date of its discovery in 1970, the 3’ poly(A) tail of eukaryotic mRNA has taken on the character of a structure in search of a function. Candidate functions favored in those far-off days were: mRNA transport from nucleus to cytoplasm, the control of mRNA stability, the compartmentalization of mRNA in the cytoplasm and the question of its association with the cytoskeleton, and, as a more remote possibility, a role in the actual mechanism or control of translation (reviewed by Brawerman, 1981). Over the intervening years, faith in the importance of the poly(A) tail as a determinant of mRNA stability has been maintained and recently rewarded by interesting and provocative results (reviewed by Bernstein and Ross, 1989). In the meantime, support for a role of the poly(A) tails and the 3’ untranslated region (UTR) in the control of translation has been more fickle, but here too there has been a recent upsurge of interest and a wave of significant new results. These new data are drawn from a wide variety of organisms and biological phenomena; this review aims to gather these different strands of evidence together, while taking care not to make unjustified extrapolations from one organism or system to another. Poly(A) Tail Synthesis and Metabolism Although the details are outside the scope of this review, a brief account of the polyadenylation process in the nucleus is probably useful. In mammalian cells the primary hnRNA transcript is cleaved at a point specified by its location within 30 nucleotides downstream of an AAUAAA motif and just upstream of a GU-rich segment that shows less sequence conservation (reviewed by Humphrey and Proudfoot, 1988). This cleavage is rapidly followed by the addition of A residues to the free 3’ end in what appears to be a two stage process. The initial stage shows a strong requirement for the AAUAAA motif, but once some ten A residues have been added, further polyadenylation seems dependent only upon the preexisting oligo(A) tail (Sheets and Wickens, 1989). Pulse-labeling experiments show that the newly synthesized mRNA detected in the nucleus carries poly(A) tails with a narrow size distribution at about 250 nucleotides in mammalian cells but nearer 100 nucleotides in lower eukaryotes. It is not clear what limits the length of the poly(A) tail, but it is unlikely to be due to an intrinsic property of the poly(A) polymerase (Sheets and Wickens, 1989). On transport to the cytoplasm, the poly(A) tail seems to undergo a slight shortening but still shows a narrow size distribution (Brawerman, 1981). Thereafter, the length becomes shorter and more heterogeneous, probably because the rate of shortening differs for different mRNA
Review
species (see below). Some quite dramatic cases of poly(A) lengthening have been observed in developmental situations, as will be discussed later, but even in somatic cells there is some end addition of A residues in the cytoplasm that partly counteracts the shortening process (Brawerman, 1981). In the presence of actinomycin to inhibit further mRNA synthesis, labeling of cytoplasmic mRNA poly(A) reaches steady-state within about 10 min, representing an average of about five A residues added to each tail. Poly(A) Binding Protein (PABP) The question of any role of the poly(A) tail in controlling mRNA stability or translation cannot be considered in isolation from the function of the poly(A) binding protein (PABP), a -70 kd protein common to most if not all eukaryotes. Cloning of the PABP genes of yeast, humans, and Xenopus has revealed a common domain structure in the N-terminal two-thirds of the protein, which consists of four repeat units, each about 90 amino acid residues in length and highly conserved in all three organisms (Adam et al., 1986; Sachs et al., 1986; Grange et al., 1987; Zelus et al., 1989). The sequence of the C-terminal one-third is conserved in humans and Xenopus, but not in yeast. In yeast, it has been possible to demonstrate that the PABP gene is essential for viability (although certain suppressor mutations can obviate the requirement). However, the complete gene can be replaced by truncated forms consisting of just a single domain with little effect on cell growth at 30% but a reduction of cell growth at 37°C (Sachs et al., 1987). Digestion of poly(A)-PABP complexes with nonspecific nucleases results in fragments with PABP bound to a length of 27 A residues (Baer and Kornberg, 1983). However, the common assumption that the minimal binding site for PABP is 27 residues may well be incorrect, as studies with poly(A) of different lengths showed that a 12 nucleotide fragment was required for efficient binding, and little increase in binding was seen with longer lengths, irrespective of whether the full-length PABP or a singledomain derivative was studied (Sachs et al., 1987). The 27 nucleotide repeat relates to the actual spacing of the PABP molecules on the poly(A) tail and not to the minimum size of the binding site. With the single domain derivative, this spacing decreases to about 13 nucleotides. Binding of PABP to poly(A) is very stable even in high salt concentrations, but this tight binding should not be regarded as immobilizing the PABP on the poly(A). There is considerable evidence that PABP molecules can migrate or transfer between poly(A) tracts, a migration that appears
to require
the
multidomain
protein
structure
and
is
not seen with the single-domain derivative (Sachs et al., 1987). It seems likely that the poly(A) tails of all cytoplasmic mRNAs in all eukaryotes are associated with PABP, but Xenopus oocytes may be an exception. lmmunoblotting methods failed to detect PABP in oocytes or early embryos, and although more sensitive methods showed that it is actually present in oocytes, the question remains
Cdl 16
open as to whether there is sufficient PABP to cover all of the mRNA poly(A) tails (Zelus et al., 1989). Maternal PABP mRNA appears to be destroyed at oocyte maturation, and the mRNA only reappears at the neurula stage of embryogenesis, the stage when PABP itself becomes easily detectable and when there is a large increase in poly(A)+ mRNA. The Role of Poly(A) Tails, 3’ UTR Sequences, and PABP in Controlling mRNA Stability A brief consideration of mRNA degradation mechanisms is necessary at this stage, since any claims for an influence of poly(A) tails and 3’ UTR sequences on translation efficiency always provoke questions as to whether the effects can be explained in terms of mRNA stability. The idea that poly(A) tail shortening and its ultimate removal is the prelude to mRNA degradation originates from the fact that the length of the tails decreases as mRNA ages in the cytoplasm, and in the case of shortlived mRNAs such as c-fos, poly(A) tail shortening is exceptionally rapid (Wilson and Treisman, 1988). It is also generally agreed that apart from species with shortened poly(A) tails, it is hard to detect any other mRNA degradation intermediates. The few exceptions are short-lived histone H4 mRNA and c-myc mRNA intermediates missing small segments of extreme 3’ UTR sequences (Ross and Kobs, 1988; Ross et al., 1986; Brewer and Ross, 1988). This suggests that once the poly(A) tail has been removed, degradation of the body of the mRNA normally proceeds rapidly and processively in the general 3’+5 direction, catalyzed either by an exonuclease or by an endonuclease that removes small fragments. For a few mRNA species, notably j3-actin mRNA, poly(A) shortening is not the prelude to rapid degradation, and there is a substantial pool of actin mRNA with, at most, a very short oligo(A) tail that is not destined for rapid destruction (Brawerman, 1981). It seems possible that these mRNAs have structures near the 3’end that make them particularly resistant to the degradative processes. Studies of mRNA degradation in a cell-free system that seems to reproduce faithfully in vivo degradation mechanisms show that the stabilizing effect of the poly(A) tail requires PABP in a way which suggests that the poly(A) tail is degraded only when it is naked and unassociated with PABP (Bernstein and Ross, 1989; Bernstein et al., 1989). This implies that the poly(A) tails of short-lived mRNAs such as c-fos and c-myc, which undergo very rapid shortening, must be naked for a higher proportion of the time than stable mRNAs such as globin. A major determinant of the rapid turnover of these mRNAs as well as lymphokine mRNAs is the multiple copies of AU-rich motifs (consensus sequence UAUUUAU) common to the 3 UTRs of these mRNAs (Caput et al., 1986; Wilson and Treisman, 1988; Jones and Cole, 1987). Transfer of these segments to the 3’ UTR of globin mRNA dramatically reduces its stability (Shaw and Kamen, 1986). It has been suggested that PABP readily migrates from the poly(A) tract to these AU-rich segments of the 3’ UTR, leaving the poly(A) tract naked and vulnerable to degradation for a high proportion of the time (Bernstein and Ross, 1989). An
alternative model postulates base pairing between the poly(A) tail and the UAUUUAU motifs in the 3’ UTR, which might exclude PABP from binding to the poly(A) tail and allow nuclease attack at the points of mismatch (Wilson and Treisman, 1988). A relationship between 3’ UTR structures and degradation is also demonstrated by the stabilization of transferrin receptor mRNA under conditions of iron deficiency, which is thought to be due to the binding of a regulatory protein to five repeated sequence motifs in the 3’ UTR (Casey et al., 1988); it is not yet clear whether this binding affects the rate of poly(A) shortening or the degradation of the body of the mRNA that follows poly(A) tail removal. However, events and sequences in other parts of the mRNA can also affect degradation rates: the selective destruction of tubulin mRNA promoted by unpolymerized tubulin subunits (Yen et al., 1988); the destabilization of c-fos mRNA promoted by a sequence element within the coding segment that acts independently of the 3’ UTR AU-rich motifs (Shyu et al., 1989); the influence of 5’-proximal sequences on c-myc degradation (Jones and Cole, 1987); the alterations in mRNA stability when the translational stop codon is moved to a different site in the mRNA (Marzluff and Pandey, 1988; Daar and Maquat, 1988); and the stabilization of some mRNAs such as c-fos and histone mRNAs that occurs when protein synthesis is inhibited by cycloheximide (Marzluff and Pandey, 1988; Wilson and Treisman, 1988). If all of these determinants influence mRNA degradation rates by affecting the rate of poly(A) shortening and/or the rate of subsequent degradation from the 3’ end of the body of the mRNA, this implies some sort of “communication” or functional interaction between the 3 end of the mRNA and far upstream segments, an interaction that could also be significant in the control of translation. The discussion above has focused on evidence from mammalian systems, and there are some doubts as to whether it can be extrapolated in its entirety to lower eukaryotes such as yeast. The short-lived mRNAs in yeast do not seem to share common 3’ UTR sequence elements such as the AU-rich motifs of mammalian lymphokine and proto-oncogene mRNAs, and for one rapidly degraded species, the MATal mRNA, the primary determinant of the instability is a short segment in the coding sequence with a high frequency of rare codons (Parker and Jacobson, 1990; Herrick et al., 1990). Moreover, there is some evidence that PABP in yeast, far from preventing poly(A) shortening, may actually promote it (Sachs and Davis, 1989). Effect of Poly(A) Tails and 3’ UTR Sequences on mRNA Stability and Ranslation in Xenopus Oocytes It is also not clear to what extent the mechanism of mammalian cell mRNA degradation is operative in Xenopus oocytes. This is an important issue since experiments involving the microinjection of poly(A)+ and deadenylated mRNAs into oocytes have provided much of the evidence on the question of a role of poly(A) tails in translation. Whereas the earlier studies used polynucleotide phosphorylase to deadenylate poly(A)+ mRNAs and poly(A)
Review: 17
Poly(A)
Tails
and 3’ UTR Sequences
polymerase for (re)addition of tails, more recent work has used in vitro transcripts of cloned genes, which has the advantage that poly(A) tails of defined length can be added, although in many such experiments the cloning strategies have given tails that are not purely poly(A). In interpreting such results it is important to distinguish between assays of functional half-life (determined by following the rate of synthesis or accumulation of the translation product) and chemical half-life (determined by following the fate of the input RNA). If polyadenylation influences both translation efficiency and mRNA stability, the functional half-life cannot be assumed to be equivalent to the chemical or structural half-life. The most influential result in this area, but unfortunately the least typical, has been the original report that the functional half-life of deadenylated globin mRNA was much shorter than that of poly(A)+ globin mRNA (Huez et al., 1974). Readenylation experiments showed that a tail of 32 A residues was sufficient to confer a long functional halflife, while addition of only 16 residues gave the same results as fully deadenylated mRNA (Nude1 et al., 1976). In addition, polyadenylation of HeLa cell poly(A)- mRNA prior to microinjection greatly increased the functional lifetime of the histone mRNAs, which was only a few hours in the absence of a poly(A) tail (Huez et al., 1978). In contrast, direct assay of the fate of the microinjected RNAs showed a much smaller difference between the stability of deadenylated globin mRNA and globin mRNA carrying either a long poly(A) tail or a 3’ tail consisting of A&& (Drummond et al., 1985; Galili et al., 1988). In similar assays with other mRNA species, the effect of polyadenylation on chemical stability over the first 24 hr following microinjection was very small, and it was only by 48 hr that significant differences were evident (Drummond et al., 1985; Galili et al., 1988). However, polyadenylation did have significant effects on translation efficiency in these experiments. The addition of long poly(A) tails increased the rate of lysozyme mRNA translation in oocytes by 20fold and chymosin mRNA translation by lO-fold (Drummond et al., 1985). A 3’ tail of A55C20 on zein mRNA stimulated translation by over 20-fold and a similar tail on j3-globin mRNA gave a 6-fold stimulation (Galili et al., 1988). All of these assays were performed at 24 hr after microinjection and took into account differences in mRNA stability, with the slight reservation that as stability is assessed by the size and amount of surviving input RNA, subtle changes in RNA structure cannot be ruled out. In experiments where the time course of protein synthesis has been examined, as in the original work with globin mRNA (Huez et al., 1974) and more recent studies using zein mRNA (Galili et al., 1988) a common feature is that the deadenylated mRNA is translated with at least 50% of the efficiency of the poly(A)+ form during the first few hours after injection. Subsequently, the efficiency of translation of the poly(A)+ mRNA seems to increase, since the mRNA moves into larger polysomes (Galili et al., 1988). In contrast, the deadenylated mRNA remains in small polysomes and its translational efficiency tends to decrease. These results go some way toward explaining the discrepancies between the functional and chemical stability
of microinjected poly(A)- mRNA. They have been interpreted in terms of a model that distinguishes between translation initiation, which is envisaged as random initiation by ribosomal subunits present in the free subunit pool, and reinitiation, where a ribosome that has just completed translation of the open reading frame has a high probability of (re)initiating translation of the same mRNA molecule. The poly(A) tail is postulated to promote the reinitiation process (as opposed to the random selection pathway), presumably through promoting some sort of interaction between 3’- and Y-proximal elements of the mRNA, such as to increase the probability that ribosomes recycle on the same mRNA (Galili et al., 1988). To what extent are these results typical of other eukaryotic cells? Electroporation has been used to introduce into plant protoplasts j3-glucuronidase (GUS) mRNA derivatives with different 5’and 3’ UTR sequences (Gallie et al., 1989). The presence of an AP5 tail increased the level of GUS activity detectable 8-24 hr after electroporation, the increase ranging from 12-fold to over 30-fold depending upon the plant species. Expanding the length of the tail to Az5GUUAz5UUUAAAGAAUU produced a further increase close to 2-fold in all species. Apparently a short length of non-poly(A) sequences downstream of the poly(A) tract was not detrimental, but longer segments of extraneous sequences largely negated the stimulatory effect of the poly(A) tract, especially in tobacco and maize. As for mRNA stability, the addition of a tail of 50 (mostly) A residues increased the half-life in tobacco protoplasts from about 30 min to 70 min, but this stabilization seems totally insufficient to account for the 30-fold increase in the yield of GUS activity. Besides the poly(A) tail itself, sequences in the 3’ UTR can also influence translation efficiency in vivo. The most striking case is that of the human p-interferon (IFN) mRNA, which is translated very inefficiently in Xenopus oocytes. Exchanging the UTR segments between different genes showed that this inhibitory effect lies mainly in the 3’UTR, with only a minor effect of the 5’ UTR. There is clearly some additional contribution from the coding sequence, since the IFN mRNA 3’ UTR transferred to the lysozyme coding sequence is significantly less inhibitory than when attached to the IFN coding sequence (Kruys et al., 1987). The negative elements in the 3’ UTR were shown to correspond to the AU-rich motifs common to lymphokine mRNAs which are thought to be determinants of the rapid turnover rates of these mRNAs in mammalian cells (Kruys et al., 1988, 1989). However, in Xenopus oocytes these sequences influence translation rather than degradation, for the microinjected IFN mRNA, whether polyadenylated or not, was stable for at least 24 hr. Moreover, the endogenous c-fos mRNA in oocytes has a half-life of many days, in contrast to its very rapid turnover in Xenopus fibroblasts (Mohun et al., 1989). It therefore seems that Xenopus oocytes may lack the mRNA degradation system found in mammalian somatic cells, or if they have such a system, then its activity would appear to be much lower and any coupling to sequences in the body of the mRNA must involve different sequence motifs.
Cdl 10
Thus, all of the experiments discussed in this section argue that in addition to any effect on mRNA stability, which seems generally small but quite variable between different mRNA species, polyadenylation can exert a direct influence on the translational efficiency of an mRNA in vivo. While this effect appears to be at the level of translation (re)initiation, it is in the nature of these assay systems that almost nothing can be ascertained as to the step in the initiation pathway that is stimulated. For evidence on this issue we must turn to other systems. The Influence of Poly(A) Tails, PABP, and 3’ UTR Sequences on Translation In Vitro Although the stimulatory effects of poly(A) tails on in vitro translation are much smaller than in intact Xenopus oocytes, such stimulation has been consistently observed in the reticulocyte lysate but not in the wheat germ system, a difference that may be related to the supposedly higher reinitiation efficiency in reticulocyte lysates. It is also of interest that the AU-rich sequence in the 3’ UTR of human p-interferon mRNA had no effect on translation in the wheat germ system but was inhibitory in the reticulocyte lysate, albeit to a lesser extent than in Xenopus oocytes (Kruys et al., 1987, 1988). In reticulocyte lysates, deadenylated ovalbumin mRNA was translated less efficiently (in competition against the endogenous globin mRNA) than the poly(A)+ form, and indirect assays showed that mRNA degradation was not the explanation for this difference (Doe1 and Carey, 1976). Deadenylation reduced elongation and initiation rates by about 20% and 40% respectively. In the nuclease-treated reticulocyte lysate, VSV mRNAs were translated at between 1.5fold and 3-fold higher efficiency if they were polyadenylated; however, neither the stability nor the ribosome transit times were affected, suggesting that differences in (re)initiation rates must be the primary cause (Jacobson and Favreau, 1983; Munroe and Jacobson, 1990). This was confirmed by studying the polysome distribution of differentially labeled poly(A)+ and poly(A)mRNAs added to the same assay: with either VSV mRNA or globin mRNA (in vitro transcripts with defined tails of pure poly(A)), the poly(A)+ form was preferentially incorporated into larger polysomes. With the globin mRNA derivatives, a tail of five A residues gave the same outcome as poly(A)- mRNA, 32 or 68 residues gave maximal stimulation, and 23 residues was intermediate (Munroe and Jacobson, 1990). When poly(A)+ and poly(A)- VSV mRNAs were compared in initiation complex formation assays, the yield of 40S-mRNA complexes (in the presence of edeine, which blocks subunit joining but also promotes aberrant scanning of 40s subunits) was the same with both species, but the yield of 80S-mRNA complexes (in the presence of anisomycin, which blocks peptide bond synthesis) with poly(A)- mRNA was only about half that obtained with the poly(A)+ form (Munroe and Jacobson, 1990). These results suggest that the absence of a poly(A) tail may reduce the rate of subunit joining, which converts a 40S-mRNA complex into an 80S-mRNA complex. However, a simple inhibition uniquely to this step should result in an accumu-
lation of 4OS-poly(A)mRNA intermediate complexes in the assays containing anisomycin. As this expectation was not fulfilled, we may need to seek a more complex interpretation. Another test for a role of poly(A) tails in translation is to determine whether free poly(A) inhibits cell-free protein synthesis. A complication here is that all homopolynucleotides inhibit at concentrations above about 30 kg/ml. However, in some circumstances, poly(A) but no other homopolymer (including poly(dA)) shows selective inhibition at much lower concentrations (5 ug/ml). This hypersensitivity was evident with poly(A) of average length of at least 51 nucleotides but not with poly(A) of 23 nucleotides. It has been seen in the nuclease-treated reticulocyte lysate or Lcell extract translating (deproteinized) poly(A)+ mRNAs, but was not seen with mRNPs (mRNA-protein complexes in which PABP is likely to be the main protein component) and nonpolyadenylated viral RNAs such as tobacco mosaic virus RNA, turnip yellow mosaic virus RNA, and reovirus mRNAs, or with endogenous mRNA in untreated cell extracts (Jacobson and Favreau, 1983; Lemay and Milward, 1986; Grossi de Sa et al., 1988; Munroe and Jacobson, 1990). Even with deproteinized globin mRNA, inhibition was only seen if the poly(A) was added at zero time rather than after a 5 min delay. It was therefore argued that poly(A) could inhibit only the translation of poly(A)+ mRNA that had not yet been assimilated into mRNA-protein complexes, presumed to be mRNA-PABP complexes. Consistent with this interpretation, the addition of PABP abolished the inhibition caused by poly(A) (Grossi de Sa et al., 1988). However, the quantity of PABP required to achieve this anti-inhibitory effect was very small, such that the model only seems self-consistent if PABP can be assumed to have a much higher affinity for the poly(A) tail of globin mRNA than for the free poly(A). When VSV poly(A)- mRNA or Dictyostelium poly(A)mRNA (mainly histone mRNA) was translated in the nuclease-treated reticulocyte lysate, or reovirus mRNA in L-cell extracts, low concentrations of poly(A) actually stimulated translation (Jacobson and Favreau, 1983; Lemay and Milward, 1986; Munroe and Jacobson, 1990). This suggests that poly(A)-PABP complex can act in trans to stimulate translation; such stimulation is presumably at the initiation step, but this remains to be proven. Taken as a whole, the in vitro evidence points to a possible role of the poly(A)-PABP complex in effecting a modest stimulation of initiation or reinitiation, which may be exerted at the subunit joining step. Control of this step of the initiation pathway has not been encountered before and was previously thought unfavorable for the “economy” of the translation machinery, since it would result in fruitless sequestration of 40s subunits and initiation factors. Genetic Evidence for a Role of PABP in Translation Initiation The role of PABP in yeast has been explored by placing the full-length PABP gene under the control of an inducible promoter or by replacing the full-length gene with a one-domain gene carrying a mutation that confers a temperature-sensitive phenotype, largely because it re-
Review: Poly(A) Tails and 3’ UTR Sequences 19
suits in reduced steady-state levels of the protein (Sachs and Davis, 1989). In both experimental systems, turning off the synthesis of PABP results in a gradual cessation of growth over about 10 hr and seemed to cause inhibition of translation initiation, although the polyribosome profiles actually show a reduction in polysome yield without the pronounced skew toward small polysomes typical of simple inhibition of initiation. As there may be avery small pool of residual PAPB, a possible explanation for these polysome profiles is that the few mRNAs still complexed with PABP are translated with normal efficiency, while the majority are translated inefficiently or not at all. Depletion of PABP also resulted in an imbalance in the levels of the accumulated free subunits, in that far more free 60s subunits than 40s were observed. The reason for this imbalance was not investigated, but if we can assume that the deficit in free 40s subunits is not due to some secondary lesion in small ribosomal subunit assembly, a possible explanation is that the “missing” 40s subunits were associated with mRNA in polyribosomes, as would occur if there were a defect in the subunit-joining step of the initiation pathway. Using the one-domain mutant, suppressors of the temperature-sensitive phenotype were isolated and further selected for cold sensitivity. These suppressors, which all allowed growth in strains with a complete deletion of the PABP gene, fell into seven complementation groups. One was shown to be a mutation in a protein of the 60s ribosomal subunit, and at least five of the other six groups can be assumed to affect the 60s ribosomal subunit since they conferred hypersensitivity to cycloheximide (Sachs and Davis, 1989). These results are provocative because they introduce a novel type of evidence to the problem and clearly show a relationship between the poly(A)-PABP complex and the role of the 60s subunit in translation. Further data delineating this relationship are eagerly awaited, but in the meantime possible explanations are that poly(A) tails uncomplexed with PABP inhibit protein synthesis initiation, perhaps at the subunit joining step, or that the poly(A) tail with bound PABP may be regarded as a positive effector of this step. Either explanation is consistent with the results obtained using the reticulocyte lysate system. The first would seem to be favored by the fact that many of the more stable mRNAs in yeast persist (and are presumably translated) for some time in the poly(A)- state (Herrick et al., 1990) but such an interpretation is probably premature as there is no information yet on whether the actual efficiency of translation of these mRNAs in vivo is affected by deadenylation. If the poly(A)-PABP complex does play a role in translation initiation in yeast, one might expect to be able to suppress PABP deficiency by mutations in the genes coding for initiation factors, but it is possible that the cold-sensitive selection step used by Sachs and Davis (1989) biased the selection in favor of suppressor mutations affecting ribosome structure and assembly. Another and surprising consequence of PABP depletion is that the length of the poly(A) tails increases from a heterogeneous size of 20-60 residues, typical of cytoplasmic mRNA in wild-type cells, to 90-100 residues, a length characteristic of nuclear mRNA (Sachs and Davis,
1989). Thus, PABP seems responsible for the shortening that normally occurs during mRNA transport from nucleus to cytoplasm. However, as the extragenic suppressors restored translation without causing shortening of the unusually long poly(A) tails, it seems unlikely that the role of PABP in the shortening process is directly linked to its role in translation. Correlations between the Polyadenylation State and the Efficiency of Translation of mRNAs during Development There are some striking examples where polyadenylation of a particular mRNA is tightly correlated with activation of its translation. Table 1 gives a representative list drawn from a wide range of organisms. Although most of these examples relate to developmental systems, there are sufficient exceptions to show that this is not a peculiarity of development. The changes in polyadenylation status are usually quite selective and far from nonspecific; in Xenopus oocyte maturation, for example, the majority of mRNAs undergo deadenylation and total cell poly(A) content declines by about 50% (Sagata et al., 1980), while a subset of mRNAs becomes polyadenylated. This polyadenylation seems to involve the addition of A residues to a preexisting short oligo(A) tail rather than de novo polyadenylation. In marine invertebrates, such as Spisula, the untranslated mRNAs in unfertilized eggs are capable of binding to poly(U)-Sepharose but not to oligo(dT)-cellulose, which implies a tail of some 10 A residues (Rosenthal and Ruderman, 1987). In higher eukaryotes the average size of the preexisting oligo(A) tails seems somewhat larger. The deadenylation of certain mRNAs, which occurs during development, generally correlates with a reduction in the rate of their translation in vivo (Table 1). There are, however, a few cases where this correlation seems to break down, as illustrated by the behavior of the Spisula 4Y4 mRNA species and of ribosomal protein mRNAs in developing Dictyostelium (Table 1). However, as there are certainly other translational control mechanisms independent of poly(A), the occurrence of some exceptions seems hardly surprising. There are also a few cases where translational activation actually correlates with deadenylation. Even during Xenopus oocyte maturation and early embryogenesis, where polyadenylation correlated with translational activation is the general rule, histone mRNAs become deadenylated at about the same time that their translation is stimulated. By the criterion of binding to oligo(dT)-cellulose, 50%-750/o of histone mRNAs are poly(A)+ in oocytes, but only about 25% are poly(A)+ in early embryos, when there is a very significant activation of translation of core histone mRNAs, followed somewhat later by activation of histone Hl synthesis (Ruderman et al., 1979; Woodland, 1980). At least in the case of histone H4 mRNA the maturation-dependent deadenylation involves complete removal of the rather short oligo(A) tails (Ballantine and Woodland, 1985). In mammalian spermiogenesis many mRNAs, including the protamine mRNAs, are stored as untranslated
oocyte
Mouse
rhythm
stimulation
preproinsulin
vasopressin 2-3
increase
x increase
unspecified
The top half lists cases reporting data on the distribution of the relevant mRNA between polysome-associated as changes in translation rate. a These mRNAs bind efficiently to poly(U)-Sepharose. implying tails of about 10 A residues. b These mRNAs bind poorly to poly(U)-Sepharose, implying tails of less than 10 A residues.
glucose
circadian
Rat insulinoma
Rat hypothalamus
increase increase
vasopressin oxytocin
osmotic
Rat hypothalamus
unspecified unspecified
little change? significant increase
30 40
1
H4 histones
x increase
x increase
6 x increase x decrease
>20
2-5
25 x increase
5 5
90 >90
10 50 0 15 90
0 0 0 75 50 0
Before
% of mRNA in polysomes
Translation
>20
Hsp90
2
phosphoribosyltransferase
plasminogen
hypoxanthine actin
tissue
1 protein
mRNAs
Ll
of Translation
Starfish (M. glacialis) oocyte maturation
Sea urchin (S. purpuratus) early embryogenesis
maturation
spermiogenesis
Mouse
protamine transition
most mRNAs ribosomal protein
switch from growth to development
Dictyostelium vegetative
protein
A9 84 D7 GIO ribosomal
Frog (X. laevis) oocyte maturation
reductase
and Activation
ribonucleotide cyclin A histone H3 a-tubulin 4YlO 4Y4
Polvadenvlation
mRNA
solidissima)
between
Clam (Spisula fertilization
I. Correlations
Organism
Table Data
100 25
100
100
100 100 100 100 5b
100 100 100 25 lob 10s
After
and nonpolysomal
60 50-75
5
0
40 80 50 100 100
108 0= 50a 100 100 100
Before
% of mRNA Bound to Oligo(dT)-Cellulose
Polyadenylation
states.
lower half shows
et al. (1987)
and Brandhorst
(1986)
and
examples
where
the data have
et al. (1986)
et al. (1988)
been given
Smith et al. (1988) Ruderman et al. (1979); Ballantine and Woodland (1985) Carrazana et al. (1986); Carter and Murphy (1989)
Standart
Bedard
Muschel
The
(1989)
Huarte et al. (1987); Vassalli et al. (1989) Paynton et al. (1988)
Kleene
Robinson
H4)
(1987)
(1988) Palatnik et al. (1984) Steel and Jacobson (1988)
McGrew et al. (1989) Hyman and Wormington
240
30
400 increase
0 (histone
and Ruderman
Dworkin et al. (1985); Dworkin Dworkin-Rash (1985)
Rosenthal
References
increases by 130 nucleotides
250 unspecified
up to 14
increase
