|Home | About | Journals | Submit | Contact Us | Français|
Recruitment of the eukaryotic translation initiation factor 2 (eIF2)-GTP-Met-tRNAiMet ternary complex to the 40S ribosome is stimulated by multiple initiation factors in vitro, including eIF3, eIF1, eIF5, and eIF1A. Recruitment of mRNA is thought to require the functions of eIF4F and eIF3, with the latter serving as an adaptor between the ribosome and the 4G subunit of eIF4F. To define the factor requirements for these reactions in vivo, we examined the effects of depleting eIF2, eIF3, eIF5, or eIF4G in Saccharomyces cerevisiae cells on binding of the ternary complex, other initiation factors, and RPL41A mRNA to native 43S and 48S preinitiation complexes. Depleting eIF2, eIF3, or eIF5 reduced 40S binding of all constituents of the multifactor complex (MFC), comprised of these three factors and eIF1, supporting a mechanism of coupled 40S binding by MFC components. 40S-bound mRNA strongly accumulated in eIF5-depleted cells, even though MFC binding to 40S subunits was reduced by eIF5 depletion. Hence, stimulation of the GTPase activity of the ternary complex, a prerequisite for 60S subunit joining in vitro, is likely the rate-limiting function of eIF5 in vivo. Depleting eIF2 or eIF3 impaired mRNA binding to free 40S subunits, but depleting eIF4G led unexpectedly to accumulation of mRNA on 40S subunits. Thus, it appears that eIF3 and eIF2 are more critically required than eIF4G for stable binding of at least some mRNAs to native preinitiation complexes and that eIF4G has a rate-limiting function at a step downstream of 48S complex assembly in vivo.
A large number of eukaryotic translation initiation factors (eIFs) have been shown to stimulate different steps in assembly of the 48S preinitiation complex (PIC) in vitro. Recruitment of Met-tRNAiMet to the 40S subunit in the ternary complex (TC) with eIF2 and GTP is promoted by eIF1, eIF1A, eIF5, and the eIF3 complex (5, 22, 24). The 43S PIC thus formed interacts with mRNA in a manner stimulated by eIF4F bound to the 5′ m7G cap and by poly(A)-binding protein (PABP) associated with the poly(A) tail, forming the 48S PIC (50). The 40S ribosome scans the mRNA until the Met-tRNAiMet base pairs with an AUG triplet, triggering GTP hydrolysis by eIF2, in a reaction stimulated by eIF5. eIF2-GDP is ejected from the ribosome, and eIF5B promotes joining of the 60S subunit with the 40S-Met-tRNAiMet-mRNA complex, leading to release of the remaining eIFs and production of the 80S initiation complex (33, 44, 58).
eIF4F is comprised of the cap-binding protein eIF4E, the ATP-dependent RNA helicase eIF4A, and the scaffold subunit eIF4G, which has separate binding domains for eIF4E, eIF4A, and PABP (19, 50). Recruitment of eIF4A as a subunit of eIF4F to the 5′ end of mRNA, and also stimulation of its helicase function by eIF4G and eIF4B, is thought to facilitate unwinding of secondary structure and thereby promote binding of the 43S PIC at the 5′ end of the mRNA and subsequent scanning. As mammalian eIF4G also contains a binding domain for eIF3, it is believed to provide a protein bridge between eIF4E and PABP, bound at the ends of mRNA, and eIF3 bound to the 40S subunit as another means of stimulating 48S PIC assembly (22, 50). In Saccharomyces cerevisiae, however, interaction between eIF3 and eIF4G has not been detected, and the eIF3-binding domain identified in mammalian eIF4G1 (28, 40) is not recognizable in the amino acid sequences of the two yeast eIF4G isoforms. It has been proposed that an eIF5-eIF4G interaction might functionally substitute for the eIF3-eIF4G interaction to promote mRNA recruitment in yeast (6).
Although biochemical studies have implicated multiple factors in 40S binding of TC and mRNA in vitro, the relative importance of these factors for PIC assembly in vivo is unclear. Previously, we presented genetic and biochemical evidence that eIF1A is required for a wild-type (WT) rate of TC recruitment in yeast cells (15, 42). One explanation for the role of yeast eIF3 in 43S assembly is prompted by its physical association with TC in a multifactor complex (MFC) that also contains eIF1 and eIF5. In the MFC, the β-subunit of eIF2 binds directly to the largest, a-subunit of eIF3 (TIF32/a) (60) and indirectly to the NIP1/c subunit of eIF3 in a manner bridged by eIF5 (2, 4, 6). Interaction of eIF5 with eIF2β enhances its binding to NIP1/c (55), and all three of these factors, plus TIF32/a, can bind directly to eIF1 (60). This network of protein-protein interactions could underlie cooperative interaction of MFC components with their independent binding sites on the ribosome (24, 34, 59). Supporting this idea, point mutations in NIP1/c (60), a combination of mutations in eIF5 and TIF32/a (41), and epitope tagging of eIF1 (54) were found to destabilize the MFC and reduce 40S binding of multiple MFC components in yeast cell extracts.
To provide a more exhaustive test of this model, and also to evaluate the relative contributions of eIF3, eIF2, and eIF5 to 43S complex assembly in vivo, we determined the effects of depleting each of these factors on association of all other MFC constituents with native PICs in cell extracts. We also asked whether depletion of MFC components, or of eIF4G, would diminish the level of native 48S PICs to explore the mechanism of mRNA recruitment in yeast cells. To address these questions, we constructed strains harboring temperature-sensitive degron (td) alleles. Each td allele encodes ubiquitin and a thermolabile dihydrofolate reductase moiety (the letter omitted in some cases) fused to the N terminus of the factor of interest, expressed from a copper-dependent promoter, and integrated into the chromosome in a manner that disrupts the resident WT allele. The td mutants also express the ubiquitin ligase UBR1 from a galactose-inducible promoter. Shifting cells from medium with copper at 25°C to medium with galactose lacking copper at 36°C represses new synthesis and triggers proteasomal degradation of the degron protein (14, 29).
We have analyzed degron mutants endowed with conditional expression of (i) the β-subunit of eIF2, (ii) both TIF32/a and PRT1/b subunits of eIF3, (iii) eIF5, and (iv) eIF4G1 in a strain lacking the eIF4G2 isoform. Our results show that depletion of eIF3 subunits, eIF2β, or eIF5 decreases 40S binding of all other MFC components, supporting a model of coupled binding by MFC components to the 40S subunit. We also provide the first in vivo evidence that eIF3 and eIF2 are critical for stable mRNA binding to 40S subunits and that eIF5 is crucial for converting 48S PICs to 80S initiation complexes. Surprisingly, we found that depletion of eIF4G led to accumulation rather than reduction in 40S-bound mRNAs. Thus, at least for some transcripts, it appears that eIF3 and eIF2 can promote mRNA recruitment in vivo independently of eIF4G and the proposed eIF3-eIF4G interaction. The fact that 48S complexes accumulate in eIF4G-depleted cells also implies that eIF4G performs a rate-limiting function downstream of 48S assembly in vivo.
Strains and plasmids used in this study are listed in Tables Tables11 and and2,2, respectively. Details of their construction and confirmation are provided in the supplemental material, as is a list of oligonucleotide primers employed (see Table S1 in the supplemental material). Yeast strains were constructed using standard techniques of yeast transformation (17, 25), gene replacement (49), plasmid shuffling (11), and degron fusion (14). Media were prepared essentially as described elsewhere (52).
Whole-cell extracts (WCEs) were separated by velocity sedimentation through sucrose gradients as described previously (41) except for the separations carried out in a high-salt buffer (16). The analysis of 40S binding of initiation factors and mRNA in HCHO cross-linked cells was carried out as described previously (41) except for Western analysis of gradient fractions and the resedimentation of 40S fractions, which were conducted as follows. For Western analysis, 6× sodium dodecyl sulfate-polyacrylamide gel electrophoresis loading buffer (375 mM Tris-HCl [pH 6.8], 12% sodium dodecyl sulfate, 30% sucrose, 0.06% bromophenol blue [sodium salt], and 30% β-mercaptoethanol) (30) was added directly to the gradient fractions, and the samples were boiled for 5 to 10 min to reverse the cross-links prior to loading on gels. For resedimentation experiments, the conventional protocol was followed except that twofold-greater A260 units of WCE were resolved on the gradient and, following the first sedimentation, the 40S fractions were pooled, diluted with 10 volumes of extraction buffer (but lacking all inhibitors), concentrated using an Amicon Ultra-4 centrifugal filter device (Millipore), and sedimented through a second sucrose gradient under the same conditions used for the first gradient separation.
Western analysis was conducted as described previously (41) except that we employed different eIF1 antibodies (61) and commercially available monoclonal hemagglutinin antibodies (Santa Cruz Biotechnology). The immune complexes were visualized using ECL Plus (Amersham) and quantified by fluorescence imaging using the Storm model 840 PhosphorImager. Northern analysis was carried out as described previously (41) except that we used 10% polyacrylamide-Tris-borate-EDTA-urea gels (Bio-Rad Laboratories). The probe against MFA2 mRNA was generated as described previously (41) by labeling a fragment containing the MFA2 open reading frame that was PCR amplified from genomic DNA using primers mMFA2 5′ and mMFA2 3′ (see Table S1 in the supplemental material).
Because deleting each gene encoding eIF5 (TIF5), eIF2β (SUI3), eIF3a (TIF32), eIF3b (PRT1), or eIF4G1 (TIF4631) (in a strain lacking eIF4G2) is lethal (22, 24), we expected to find that depletion of the degron-tagged forms of these factors would greatly impede cell growth. Indeed, all td mutants grew poorly under restrictive conditions (Fig. 1A and B), and this growth defect was complemented by the cognate WT allele on a plasmid (data not shown). Western analysis of WCEs showed that the majority of each degron protein was eliminated after 2 h under nonpermissive conditions (Fig. (Fig.1C);1C); however, the degron proteins were still detectable in the 40S fractions after 3 h (data not shown). Based on this last finding and the fact that the growth rates of certain degron mutants did not reach their lowest levels until after ~10 h (Fig. (Fig.1B),1B), we cultured cells overnight (~16 h) under nonpermissive conditions to ensure complete depletion of each protein. Under these conditions, the growth defects were still reversible on shifting the td mutants back to permissive conditions (data not shown). The persistence of a substantial amount of 40S-associated degron-tagged eIFs after 3 h of depletion when most of these proteins have been degraded may indicate that ribosome association affords degron-tagged proteins a measure of protection from proteasomal degradation.
The sui3-td strain and the tif32-td prt1-td double mutant exhibited nearly complete loss of polysomes and accumulation of 80S monosomes under nonpermissive conditions, indicating a profound reduction in translation initiation (Fig. 2A, B, and D). The tif5-td and tif4631-td tif4632Δ strains also displayed strong, albeit less extreme, polysome runoff (Fig. 2C and E). The tif5-td and tif4631-td tif4632Δ strains contained a “half-mer” shoulder on the 80S peak that was usually larger in the tif5-td mutant (Fig. 2C and E). Half-mers are formed by mRNAs containing an elongating 80S ribosome and a 43S complex bound to the mRNA leader (i.e., a 48S complex). While the half-mer shoulders in the tif5-td and tif4631-td tif4632Δ mutants are relatively small, this would be expected from the fact that most 80S ribosomes in the monosome peak are 80S couples lacking mRNA and thus incapable of forming half-mers. (We verified that the large 80S peaks in the tif5-td and tif4631-td tif4632Δ extracts are largely 80S couples by showing that they dissociate into free 40S and 60S subunits in the presence of a high salt concentration [see Fig. S1 in the supplemental material]. We presume that half-mers are not visible on the disome peak in each of these mutants because it is much smaller and proportionally broader than the monosome peak.) Several lines of evidence indicating that the half-mer shoulders represent 48S PICs accumulating in these mutants will be presented below. The appearance of half-mers in eIF5-depleted cells is consistent with the role of this factor in activating GTP hydrolysis by eIF2 prior to 60S subunit joining. The occurrence of half-mers in the tif4631-td tif4632Δ strain (hereafter called simply tif4631-td) suggests that eIF4G also carries out a function required to convert 48S PICs into 80S complexes.
To measure the amounts of factors and mRNA bound to native PICs after depleting each factor, cells were treated with formaldehyde before harvesting to minimize dissociation of PICs during sedimentation through sucrose gradients. The levels of eIFs were measured by immunoblot analysis, and the amounts of tRNAiMet and RPL41A mRNA in the gradient fractions were assayed by Northern analysis. (The small size of RPL41A mRNA allows resolution of 48S complexes from mRNP particles containing this transcript .) As shown in Fig. Fig.3A3A (lanes 9 to 11), significant proportions of eIFs 2, 3, 5, 1, and 1A in WCEs from cross-linked WT cells cosedimented with the 40S subunit protein RPS22. We showed previously that nearly all of these factors are lost from the 40S fractions when WCEs from non-cross-linked cells are separated in sucrose gradients without the addition of heparin as a stabilizing agent (2). There are also substantial amounts of eIFs in the fractions just preceding the 40S fractions, which contain the MFC. In the case of eIF5 and eIF1, this phenomenon can obscure the 40S peak and make it difficult to assess the true amounts of these factors bound to 40S subunits. As described below, we took several steps to overcome this limitation.
Simultaneous depletion of the TIF32/a and PRT1/b subunits of eIF3 in the tif32-td prt1-td double mutant produced nearly complete loss of all five essential eIF3 subunits from both the 40S fractions and WCEs at large (Fig. (Fig.3A,3A, cf. input lanes and fractions 9 to 11 in the right and left panels). Thus, NIP1/c, TIF34/i, and TIF35/g eIF3 subunits are unstable in the absence of the other two core eIF3 subunits, meaning that tif32-td prt1-td cells exhibit conditional expression of the entire core eIF3 complex. We consistently observed reductions in the amounts of eIF2 subunits, tRNAiMet, eIF5, and eIF1 in the 40S fractions of the tif32-td prt1-td mutant compared to the WT strain (Fig. 3A and B). The levels of these factors in fractions 9 to 11 were measured in three to eight independent experiments, and the mean values with standard errors, expressed as percentages of the WT values determined in parallel, are given in Fig. Fig.3E.3E. We observed comparable reductions in 40S-associated eIF2α, eIF2γ, and tRNAiMet in the eIF3-depleted cells (and in other degron mutants described below); hence, the percentages of WT binding measured for these three factors were combined and averaged to provide a composite determination of 40S-associated TC. The results of these measurements suggest that cells depleted of eIF3 show a ~50% reduction in 40S binding of TC and greater reductions (of ~75%) in 40S binding by eIF5 and eIF1. Note that very similar levels of the 40S subunit protein RPS22 were observed in the 40S fractions of mutant and WT extracts (Fig. 3A and E).
In an effort to substantiate our conclusion that depletion of eIF3 reduces 40S binding of TC, eIF5, and eIF1, we calculated the distributions of these factors across the gradient by quantification of Western data from a representative experiment. As shown in (Fig. 4A, B, D, and E), only the 40S forms of these factors were substantially reduced by depletion of eIF3, with little or no reduction seen in other fractions of the gradient. Thus, depletion of eIF3 elicits a redistribution of TC, eIF5, and eIF1 from 40S complexes to more slowly sedimenting forms of these factors. The fact that we did not observe accumulation of TC, eIF5, and eIF1 at the top of the gradient for the eIF3-depleted cells probably indicates that the unbound factors in the mutant extract are less stable during centrifugation, as only eIF1 and tRNAiMet showed a reduction in the input sample prior to centrifugation (Fig. 3A and B, “In” lanes). We previously observed a similar degradation of unbound factors in cross-linking analysis of point mutations in eIF3 subunit NIP1/c (61). In contrast to the behavior of TC, eIF5, and eIF1, there was only a slight reduction in the 40S signal for eIF1A, which was smaller in magnitude than the decrease in eIF1A abundance seen in the upper fractions of the gradient for the eIF3-depleted extract (Fig. (Fig.4F).4F). Hence, we cannot judge whether eIF3 depletion impairs 40S binding by eIF1A or merely decreases the stability of eIF1A throughout the gradient.
In a second approach to confirm our conclusion that depleting eIF3 reduces 40S association of TC, eIF5, and eIF1, we resedimented the 40S fractions collected from gradients of the type shown in Fig. Fig.3A3A and quantified the levels of eIFs that cosedimented a second time with the 40S subunits. As shown in Fig. Fig.5A,5A, we observed well-defined 40S peaks for all of the eIFs after resedimenting the 40S fractions from WT cells. Importantly, resedimentation of the 40S fractions from eIF3-depleted cells revealed reductions in 40S-bound TC, eIF5, and eIF1 (Fig. 5B and D) comparable in magnitude to those measured with the conventional technique using a single sedimentation step (Fig. 3A and E). These results support our conclusion that depletion of eIF3 reduces the 40S association of all other components of the MFC.
In some resedimentation experiments, we observed proportions of eIFs in the upper fractions of the gradient for both WT and eIF3-depleted extracts (Fig. 5A and B). At least in the case of WT extracts, this would be expected if the 40S fractions from the first gradient were contaminated with unbound MFC. However, we cannot exclude the possibility that PICs are incompletely cross-linked in vivo and that the reduced 40S association of TC, eIF5, and eIF1 seen in eIF3-depleted extracts results from a higher dissociation rate of these factors from the unstable mutant PICs during the centrifugation steps. To the extent that this occurs, it would indicate that depletion of eIF3 decreases the stability of PICs, rather than reducing their formation in vivo. In any event, it is clear from our results that eIF3 is required for optimal interaction of the other three MFC components with native PICs. (Henceforth, we use the term 40S binding to signify the steady-state level of 40S interaction measured in extracts, not the rate of PIC assembly in vivo.)
Mammalian eIF3 is thought to bridge the interaction between 40S ribosomes and the eIF4G subunit of eIF4F to facilitate mRNA recruitment (22, 51). Accordingly, we expected to observe reduced amounts of RPL41A mRNA in the 40S fractions on depletion of eIF3. Although this was not observed in the tif32-td prt1-td gradients (Fig. (Fig.3B),3B), we questioned this conclusion because a distinct peak of mRNA was not present in the 40S fraction of eIF3-depleted cells, whereas a 40S peak of mRNA was observed in the WT extract (Fig. (Fig.3B).3B). Due to the dissociation of polysomes on depletion of eIF3 (Fig. (Fig.2D),2D), there are higher levels of mRNA in all fractions above the 40S region towards the top of the gradient. We reasoned that this effect could obscure a reduction in mRNA binding in the tif32-td prt1-td extract due to cosedimentation of unbound mRNPs with free 40S subunits, and we took several approaches to address this possibility.
First, we accentuated the 40S mRNA peak in the tif32-td prt1-td extract by analyzing strains lacking RPL11B, one of two genes encoding the 60S protein Rpl11p. The absence of RPL11B in the WT strain reduced the steady-state level of 60S subunits and produced half-mers on the 80S and polyribosome peaks (data not shown), as expected from a reduced rate of 40S-60S joining (48). Consistent with this, deletion of RPL11B increased the mRNA level in the 40S region by ~1.7-fold due to accumulation of free 48S complexes (Fig. (Fig.3C).3C). Second, we resedimented the 40S fractions collected from gradients prepared from rpl11bΔ derivatives of the WT and tif32-td prt1-td strains and measured RPL41A mRNA across the gradient by Northern analysis. As shown in Fig. Fig.3D3D (and quantified in panel F), we now saw a much smaller 40S mRNA peak for the tif32-td prt1-td rpl11bΔ strain, only ~13% of that seen in the WT rpl11bΔ strain, plus an accumulation of unbound mRNA near the top of the gradient. We believe that the unbound mRNA derives from the abundant free mRNP produced by polysome runoff that contaminates the 40S fractions obtained from the first separation of the mutant extract. However, it could result instead from a higher rate of dissociation of unstable 48S complexes lacking eIF3, as suggested above for unbound eIFs observed in resedimentation experiments. Hence, the reduction in 40S-bound mRNA in eIF3-depleted extracts shows that elimination of eIF3 decreases the formation of 48S PICs in vivo or their stability in extracts.
The depletion of eIF2β in the sui3-td strain nearly eliminated 40S binding by eIF2α, eIF2γ, and tRNAiMet (Fig. 6A, B, and D). Examination of the input signals in Fig. 6A and B showed that the levels of these TC components in the starting extract were reduced by eIF2β depletion. Nevertheless, the distributions of all three factors were shifted from the 40S region to the upper portions of the gradient (Fig. 4A and B and data not shown). Thus, depletion of eIF2β produced the expected strong reduction in 40S binding of other TC components (to ~10% of WT) (Fig. (Fig.6D).6D). It also led to a moderate (~40%) reduction in 40S binding of all eIF3 subunits without decreasing the abundance of these proteins in the WCE (Fig. 6A and D). Much larger decreases in 40S binding of eIF5 and eIF1 were observed, falling to only ~20% of the WT levels (Fig. 6A and D). In addition, the distributions of eIF1 and eIF5 were shifted from the 40S fractions to the top of the gradient (as was eIF3) in this degron mutant (Fig. 4C to E). We confirmed the defects in 40S binding by eIF3, eIF5, and eIF1 in the sui3-td mutant by resedimentation analysis (Fig. 5C and E). Although the level of 40S binding by eIF1A was reduced to ~60% of WT (Fig. 6A and D), we did not observe an obvious redistribution of eIF1A from the 40S fractions to the top of the gradient (Fig. (Fig.4F).4F). Thus, we conclude that depletion of eIF2β leads to strong reductions in 40S binding of eIF5 and eIF1 and a moderate reduction in 40S-bound eIF3 and has an indeterminate effect on eIF1A recruitment.
There is biochemical evidence that TC is a prerequisite for stable binding of mRNA to the 43S complex in vitro (9, 36, 57); thus, it was important to determine whether depletion of eIF2β would reduce 40S binding of mRNA in vivo. As in the case of eIF3 depletion, the total mRNA signal in the 40S fractions was not diminished by depletion of eIF2β, but a clear 40S peak of mRNA was not observed (Fig. (Fig.6B).6B). Importantly resedimentation of the 40S fractions showed a strong reduction in 40S-bound mRNA in rpl11bΔ sui3-td cells lacking eIF2β (Fig. (Fig.6C),6C), decreasing it to 35% of the WT level (Fig. (Fig.6E).6E). Thus, TC is required for efficient mRNA binding to native 43S complexes.
Depletion of eIF5 in the tif5-td mutant led to marked reductions in 40S binding of eIF3 and eIF2 subunits, and also eIF1A, without affecting the steady-state levels of these proteins in the WCE (Fig. 7A, B, and D). The 40S binding of eIF1 and tRNAiMet was also impaired by depletion of eIF5. Although the input levels of eIF1 and tRNAiMet were reduced by eIF5 depletion (particularly eIF1) (Fig. (Fig.7A),7A), examining the distributions of these factors in the gradients revealed a significant redistribution of eIF1 and tRNAiMet from the 40S region to the upper fractions (Fig. 4B and D). The distributions calculated for eIF2, eIF3, and eIF1A also revealed specific depletion of the 40S pools of these factors in cells lacking eIF5 (Fig. 4A, C, and F). We confirmed that depletion of eIF5 reduces 40S binding of TC, eIF3, eIF1, and eIF1A by the resedimentation experiment shown in Fig. 5F to H.
Note that depletion of eIF5 is the only case where we observed a decrease (~30%) in the level of the 40S subunit protein RPS22 in the 40S fractions of the gradient (Fig. 7A and D), in accordance with the diminished 40S signal in the polysome profile shown for this mutant in Fig. Fig.2E.2E. Evidence presented below indicates that the reduced level of free 40S subunits likely results from accumulation of 40S subunits in the half-mer shoulder on the 80S peak in eIF5-depleted cells. Accordingly, we extended our Western analysis to include these 80S fractions and observed no increase in levels of eIF3, TC, eIF1A, and eIF1 in the half-mer fractions of the tif5-td extract above those observed in the corresponding fractions of the WT extract (data not shown). Hence, the depletion of eIFs in the 40S fractions of eIF5-depleted cells does not result from a shift to the 80S half-mer.
It is striking that the level of 40S-bound mRNA increased ~4-fold on depletion of eIF5 in the tif5-td mutant. This effect was obvious in experiments conducted either without (Fig. (Fig.7B)7B) or with (Fig. 7C and E) resedimentation of the 40S subunits. These results support the idea that eIF5 performs a rate-limiting function required for 60S subunit joining by stimulating the GTPase activity of eIF2 at the start codon. In this view, 48S complexes accumulate and produce the increased association of RPL41A mRNA with free 40S subunits that we observe on depletion of eIF5. (Below, we discuss how eIF5 depletion can lead simultaneously to increased 40S binding of mRNA but decreased 40S binding of MFC components.)
If the accumulation of 40S-bound mRNA that occurs on depletion of eIF5 represents the build-up of functional 48S PICs incapable of subunit joining, it should depend on both eIF2 and eIF3, as these factors were shown above to be necessary for strong mRNA binding to 40S ribosomes. To test this prediction, we constructed double mutants in which TIF32/eIF3a or eIF2β was depleted simultaneously with eIF5. As shown in Fig. 8A to D, depletion of eIF3 or eIF2 subunits suppressed the accumulation of 40S-bound mRNA produced by depleting eIF5 alone. Codepletion of eIF2 or eIF3 also reduced the size of the half-mer shoulder on the 80S peak and suppressed the depletion of free 40S subunits in eIF5-depleted cells (Fig. 8E and F). (Note that the experiments in Fig. Fig.88 involved a single round of velocity sedimentation, as accumulation of 40S-bound RPL41A mRNA is readily observed on depletion of eIF5 in otherwise-WT tif5-td cells.) Together, these findings demonstrate that the accumulation of 48S complexes resulting from depletion of eIF5 depends on the upstream functions of eIF2 and eIF3 in mRNA recruitment. They also validate our interpretation of the half-mer shoulder on the 80S peak, as the accumulation of 48S PICs containing many different mRNAs (in addition to RPL41A) in eIF5-depleted cells.
Given the role of eIF4G as the scaffold subunit of eIF4F and its postulated function as an adaptor between the 40S/eIF3 complex and mRNA-binding factors eIF4E and PABP, we expected that depletion of eIF4G would produce a strong reduction in mRNA binding to free 40S subunits, comparable to what occurs in eIF3-depleted cells. We did observe a higher level of unbound RPL41A mRNA associated with polysome runoff on depletion of eIF4G in both RPL11B and rpl11bΔ strains. However, the level of RPL41A mRNA in the 40S fractions was ~1.5-fold higher in cells lacking eIF4G versus WT and, importantly, a 40S mRNA peak was clearly evident in the eIF4G-depleted extracts (Fig. 9B and C). A similar finding was made for a second small mRNA, MFA2 (Fig. (Fig.9B,9B, middle panel). Moreover, resedimentation of 40S fractions from the WT and tif4631-td strains containing rpl11bΔ confirmed that depletion of eIF4G leads to an ~1.5-fold accumulation of 48S complexes containing RPL41A mRNA (Fig. 9D and F). These last findings are consistent with the appearance of a half-mer shoulder on the 80S peak in the tif4631-td extract (Fig. (Fig.2C),2C), suggesting that many different cellular mRNAs accumulate in 48S PICs on depletion of eIF4G. The accumulation of 48S PICs in tif4631-td cells suggests that eIF4G is less critical than eIF3 and eIF2 for assembly of 48S PICs in vivo, as 40S-bound mRNA was depleted in cells lacking eIF2 or eIF3 (Fig. (Fig.3F3F and and6E).6E). It further suggests that eIF4G performs an important function downstream of mRNA binding, such that 48S complexes are not efficiently converted to translating 80S ribosomes in cells lacking eIF4G.
Finally, there were either small (~20%) or no significant reductions in 40S binding of TC, eIF5, eIF3, eIF1, or eIF1A on depletion of eIF4G, whether measured in the conventional assay (Fig. 9A, B, and E) or by resedimentation analysis of 40S-bound factors (see Fig. S2 in the supplemental material). Thus, as expected, eIF4G is not required for a WT level of 43S PIC assembly in vivo.
We have employed degron mutants to determine the effects of depleting various eIFs on the rates of translation initiation in vivo and the composition of native 43S/48S PICs. We found that simultaneous depletion of eIF3 subunits TIF32/a and PRT1/b eliminated the entire eIF3 complex and reduced the 40S binding of TC, eIF5, and eIF1 in extracts of HCHO cross-linked cells. Depletion of eIF2β, in addition to having the expected deleterious effect on TC recruitment, lowered the levels of 40S-bound eIF5, eIF1, and eIF3. Likewise, depletion of eIF5 reduced the 40S binding of eIF3, TC, and eIF1. These findings indicate that components of the MFC are interdependent for optimal binding to native PICs. At the same time, the fact that significant 40S binding of the remaining MFC components was retained in cells lacking eIF2, eIF3, or eIF5 shows that these MFC components are not completely dependent on one another for 40S binding in vivo. Results of in vitro experiments using purified mammalian or yeast factors indicate that eIF3 stimulates 40S association of TC and eIF1 (12, 24, 36, 39), that TC enhances 40S association of eIF3 (27, 57), and that eIF1 stimulates 40S binding of TC (1, 39). Thus, interdependency among MFC components in 40S binding is now firmly established for both reconstituted and native PICs. This interdependency could reflect cooperative binding, wherein physical contacts between MFC components mutually increase their affinities for independent binding sites on the ribosome. Alternatively, binding of one component of the MFC could alter the structure of the 40S subunit to produce a higher-affinity binding site for another MFC constituent.
We observed that depletion of eIF2β and eIF5 led to reductions in 40S binding by eIF1A (Fig. (Fig.66 and and7).7). In the case of eIF2β depletion, this might only reflect the decreased total amount of eIF1A present in the mutant extract. This was not the case for eIF5 depletion, however, indicating a novel role for eIF5 in recruitment of eIF1A.
Apart from its important role in 43S assembly, we found that eIF5 is required to convert 48S complexes to 80S initiation complexes in vivo. Thus, depletion of eIF5 produced a marked increase in 40S-bound mRNA and a half-mer shoulder on the 80S peak. This phenotype is consistent with the role of eIF5 as GTPase-activating protein for eIF2, as loss of eIF5 should elicit accumulation of 48S complexes stalled at the AUG codon. Our results are significant because, in previous studies, depletion of yeast eIF5 (38) or temperature-sensitive lethal mutations in eIF5 that impaired its GTPase-activating function in vitro (37) led to polysome runoff without the appearance of a half-mer, and the binding of mRNA to 40S subunits was not examined.
It is thought that hydrolysis of GTP in the TC is a prerequisite for dissociation of other eIFs from the 48S complex following AUG recognition. Thus, one might expect to observe accumulation of eIFs in parallel with mRNA on free 40S subunits on depletion of eIF5, whereas we found reductions in 40S-bound eIF3, TC, eIF1A, and eIF1 in the tif5-td mutant (Fig. (Fig.7D7D and and5F).5F). To account for this finding, we propose that depleting eIF5 has a more severe effect on 40S-60S subunit joining than on assembly of 43S and 48S PICs, leading to accumulation of 48S complexes stalled at the AUG start codons in mRNA. If the production of 48S PICs by mRNA binding to 43S complexes normally occurred more slowly than the conversion of 48S to 80S complexes, then the relatively high steady-state level of free 43S intermediates could be reduced by eIF5 depletion, but the relatively rare 48S PICs could accumulate when their conversion to 80S complexes was completely blocked by the absence of eIF5. Another possibility is that the other eIFs tend to dissociate from 48S complexes lacking eIF5 that are stalled at AUG codons (either in the cell or in extracts during sedimentation), leaving mRNA bound to partly disassembled 48S complexes. This last possibility is consistent with the fact that we did not observe accumulation of eIFs in the 80S half-mer, which by definition contains 48S but not 43S complexes, in extracts of eIF5-depleted cells.
In the current model for 40S binding of mRNA in mammalian cells, eIF3 plays an important role as an adaptor between the 40S subunit and eIF4G. This model is based largely on physical interaction between eIF3 and an internal segment of eIF4G in mammals (26, 28, 31). We demonstrated that eIF3 depletion strongly reduces the level of 40S-bound RPL41A mRNA (Fig. 3D and F) and also diminishes the 80S half-mer in eIF5-depleted cells (Fig. (Fig.8E),8E), showing that eIF3 is crucial for stable binding of many, if not all, mRNAs to native 43S PICs. Previously, we reported that the prt1-1 point mutation in eIF3b produces an accumulation of 40S-bound mRNA and the appearance of half-mers on the 80S peak in vivo (41). Thus, whereas prt1-1 impairs a function of eIF3 downstream of 48S complex formation, leading to accumulation of 48S PICs, the complete elimination of eIF3 impedes 48S assembly in vivo.
Loss of eIF2β also impaired binding of RPL41A mRNA (Fig. 6C and E) and many other cellular mRNAs (Fig. (Fig.8F),8F), providing the first demonstration that 40S-bound TC is a prerequisite for efficient mRNA binding to PICs formed in vivo (8, 36, 56). The contribution of TC to mRNA binding could reflect the stabilizing effect of base pairing between Met-tRNAiMet and the AUG codon (36, 45) or the RNA binding activity of eIF2β (32), but since depletion of eIF2β reduces 40S binding of eIF3, it could also diminish mRNA binding by decreasing eIF3 recruitment.
Contrary to expectations, depletion of eIF4G did not reduce the amounts of RPL41A and MFA2 mRNAs bound to native PICs, leading instead to moderate accumulation of free 48S complexes containing these transcripts (Fig. (Fig.9).9). We also consistently observed an 80S half-mer in the eIF4G-depleted cells (Fig. (Fig.2C),2C), suggesting that a sizable population of different mRNAs accumulates in 48S PICs in the absence of eIF4G. These findings suggest that, in contrast to eIF3 and eIF2, eIF4G is not essential for efficient recruitment of many mRNAs in vivo. However, there are several caveats to this conclusion that must be considered.
One caveat is that a very small proportion of eIF4G remaining in the eIF4G-depleted cells that is undetectable by Western analysis might be sufficient for a wild-type level of eIF4G activity in mRNA recruitment. At odds with this possibility is the fact that eIF4G1 is one of the least abundant initiation factors in yeast (62). More importantly, our depletion of eIF4G was effective enough to arrest cell growth and greatly impair translation initiation and, in the process, elicit accumulation of mRNA on free 40S subunits and a half-mer shoulder on the 80S peak. Thus, even if eIF4G were not fully depleted, the strong reduction in translation initiation and buildup of 48S complexes in the tif4631-td mutant suggest that eIF4G has a rate-limiting function downstream of mRNA binding to the 40S subunit.
It could also be argued that the accumulation of RPL41A mRNA at the top of the gradient in eIF4G-depleted cells (Fig. (Fig.9)9) indicates that mRNA recruitment is, in fact, strongly impaired in the absence of eIF4G. While we cannot dismiss this possibility, we think it is unlikely that there are sufficient 43S complexes in the cell to assemble 48S PICs on the entire cellular population of mRNAs that dissociate from polysomes in eIF4G-depleted cells. In this view, the accumulation of free mRNP is the inevitable consequence of polysome runoff regardless of the rate of mRNA recruitment.
Another possibility is that RPL41A, MFA2, and the population of mRNAs that accumulate in the 80S half-mer found in eIF4G-depleted cells represent a proportion of total yeast mRNAs that do not require eIF4F function for mRNA recruitment because their leader sequences contain little secondary structure. Indeed, Pestova and Kolupaeva reported that 48S PICs can be formed in the absence of eIF4F and ATP in a reconstituted system using an artificial mRNA lacking all secondary structure in its leader (43). Further experiments will be undertaken to test this explanation of our findings.
Finally, it is possible that the rates of 40S binding by RPL41A, MFA2, and other mRNAs that are seemingly eIF4G independent are actually reduced in the absence of eIF4G, but this rate reduction is masked by a more severe defect downstream of 48S PIC assembly, resulting in the net accumulation of 48S complexes that we observed in eIF4G-depleted cells. The central domain of mammalian eIF4G (which binds eIF4A), along with eIF4A and eIF4B, has been implicated in ATP-dependent enhancement of ribosomal scanning past secondary structures in vitro (43). Furthermore, a segment of mammalian eIF4G located upstream of the eIF4A-binding site, which contains general RNA-binding activity, was shown to be required for scanning after 40S binding to internal ribosome entry sites in certain viral mRNAs (47). Yeast eIF4G also interacts with eIF1 and contributes to stringent AUG selection in vivo (21). Thus, it is possible that 43S PICs can bind efficiently to the 5′ ends of many mRNAs in the absence of eIF4G but cannot scan effectively to the start codon. The fact that depletion of eIF4G produced a smaller accumulation of 40S-bound mRNA (~1.5-fold) than did depletion of eIF5 (~4-fold) may indicate that such 48S PICs arrested at sites of secondary structure in the leader are less stable than those which reach the unstructured initiation region. It is also possible that the putative defect in scanning in eIF4G-depleted cells is not as severe as the block to subunit joining in eIF5-depleted cells, allowing a somewhat higher rate of 48S-80S complex conversion in cells lacking eIF4G versus eIF5.
In contrast to the mammalian system, stable interaction between eIF3 and eIF4G has not been reported for the yeast factors, which might indicate a more prominent role for an eIF4G-independent pathway of mRNA recruitment in yeast. Because TC stimulates mRNA binding to the 40S subunit (9, 36, 56), eIF3 could act indirectly to promote mRNA binding by enhancing TC recruitment. Interaction between eIF4B (encoded by TIF3) and TIF35/eIF3g (63) could also be involved. Mammalian eIF3 contains subunits that bind RNA as isolated proteins (including the homologs of yeast TIF32/a and TIF35/g) (3, 7, 10, 20, 64), such that eIF3 could interact directly with mRNA in the initiation complex. Indeed, the conserved a, b, and c subunits of mammalian eIF3 are cross-linked to β-globin mRNA in 48S preinitiation complexes (27, 64), and oligonucleotides stimulate eIF3 binding to 40S subunits in a manner that correlates with their ability to bind ribosomes (27). Hence, eIF3 may stimulate mRNA recruitment by directly interacting with mRNA in addition to providing a protein bridge between eIF4G and the 40S subunit.
We thank Tom Dever for critical reading of the manuscript, Christie Fekete and Ernest Hanning for antibodies, and members of our laboratory for many helpful suggestions.
This work was supported in part by the Intramural Research Program of the NIH, NICHD.
†Supplemental material for this article may be found at http://mcb.asm.org/.