Intracellular translation initiation factor levels in Saccharomyces cerevisiae and their role in cap-complex function


  • Tobias von der Haar,

    Corresponding author
    1. Posttranscriptional Control Group, Department of Biomolecular Sciences, UMIST, PO Box 88, Manchester M60 1QD, UK.
    • For correspondence. E-mail Tobias.; Tel. (+44) 161 2008916; Fax (+44) 161 2008918.

    Search for more papers by this author
  • John E. G. McCarthy

    1. Posttranscriptional Control Group, Department of Biomolecular Sciences, UMIST, PO Box 88, Manchester M60 1QD, UK.
    Search for more papers by this author


Knowledge of the balance of activities of eukaryotic initiation factors (eIFs) is critical to our understanding of the mechanisms underlying translational control. We have therefore estimated the intracellular levels of 11 eIFs in logarithmically growing cells of Saccharomyces cerevisiae using polyclonal antibodies raised in rabbits against recombinant proteins. Those factors involved in 43S complex formation occur at levels comparable (i.e. within a 0.5- to 2.0-fold range) to those published for ribosomes. In contrast, the subunits of the cap-binding complex eIF4F showed considerable variation in their abundance. The helicase eIF4A was the most abundant eIF of the yeast cell, followed by eIF4E at multiple copies per ribosome, and eIF4B at approximately one copy per ribosome. The adaptor protein eIF4G was the least abundant of the eIF4 factors, with a copy number per cell that is substoichiometric to the ribosome and similar to the abundance of mRNA. The observed excess of eIF4E over its functional partner eIF4G is not strictly required during exponential growth: at eIF4E levels artificially reduced to 30% of those in wild-type yeast, growth rates and the capacity for general protein synthesis are only minimally affected. This demonstrates that eIF4E does not exercise a higher level of rate control over translation than other eIFs. However, other features of the yeast life cycle, such as the control of cell size, are more sensitive to changes in eIF4E abundance. Overall, these data constitute an important basis for developing a quantitative model of the workings of the eukaryotic translation apparatus.


In yeast, as in eukaryotes in general, the process of translation relies on the activities of translation initiation factors or eIFs for both facilitating and controlling the access of ribosomes to mRNAs. In Saccharomyces cerevisiae, at least 24 polypeptides are essential for the cytoplasmic assembly of ribosome–mRNA complexes that are competent for elongation (McCarthy, 1998; Hershey and Merrick, 2000).

Non-translating ribosomal subunits occur in 80S particles, from which the small subunit has to be dissociated and prepared for recruitment of the mRNA (Chaudhuri et al., 1999). This involves the assembly of a large protein complex on the surface of the 40S subunit, resulting in the formation of a 43S particle. The eIFs 1, 1A, 2, 3 and 5 have been implicated in this process. Of these factors, eIF2 and eIF3 are multisubunit complexes (Asano et al., 1998; Kimball, 1999), and eIF2 functions in translation as a ternary complex with GTP and inline image. Although all of these factors appear to be required for formation of the 43S particle, not all of them are intrinsic components of the 43S complex (Chaudhuri et al., 1997).

The eIF4 group of initiation factors mediates recruitment of this large complex to the mRNA. In the most common form of translation, the cap-binding protein, eIF4E, binds the m7GpppX structure present at the 5′ end of cellular transcripts. This protein associates with eIF4G, which in turn possesses a number of binding sites for other factors, thus acting as a nucleation site for additional activities. Among the proteins bound by eIF4G are the helicase eIF4A (Etchison and Milburn, 1987), thought to be required together with its cofactor eIF4B for the unwinding of secondary structure from the 5′ end of the message, and Pab1 (Tarun and Sachs, 1996). The binding of Pab1 is thought to mediate the synergistic effect that the 5′-cap and 3′-poly(A) tail appear to have on translation. Contacts between the cap-binding complex and the 43S-bound initiation factors are believed to be responsible for the recruitment of the 40S subunit onto the mRNA (Etchison et al., 1982; Korneeva et al., 2000; Pestova et al., 2001).

Following mRNA recruitment, the 40S subunit translocates in the 5′→3′ direction along the message. A large body of evidence points to the importance of this scanning process for translation (Kozak, 1989; Pestova and Hellen, 1999). However, neither the mode of movement nor the source of the motile force are at present understood. Scanning is terminated with the recognition of an AUG codon (Kozak, 1978) via base-pairing with the anticodon of the eIF2-GTP-inline image ternary complex. The codon-anticodon contact is thought to trigger eIF5-promoted hydrolysis of the GTP (Das and Maitra, 2001), resulting in ejection of the bound eIFs, and joining of the large ribosomal subunit in a reaction catalysed by the eIFs 5A and 5B (Pestova et al., 2000).

Mechanisms by which eukaryotic cells exert rate control on the initiation process described above rely either on cis-acting RNA features or on modifications of the canonical initiation factors. The mRNA features that are employed in the control of translation are specialized sequence elements which are often, but not always, situated in the 5′ UTR. These elements impede or enhance the process described above, either on their own or in concert with trans-acting factors, like the mammalian iron responsive element and iron regulatory protein (Munro, 1990; McCarthy, 1998).

The mechanisms responsible for modulating canonical eIF activities include phosphorylation (Flynn and Proud, 1995), changes in protein levels, and activation of competitive binding partners which sequester proteins from the complexes in which they are active. An example of the latter type of mechanism is control via the mammalian 4E-BP proteins, which prevent association of eIF4E with eIF4G, thereby interfering with the mRNA recruitment step (Haghighat et al., 1995). Initiation factor modifications can be employed by the cell for global translational control, but they can also achieve significant transcript-specific effects. For example, translation of the yeast GCN4 transcript is exceptionally sensitive to the availability of active eIF2 (Hinnebusch, 1997). Similarly, small changes in eIF4E activity in S. cerevisiae have been shown to repress translation of the CLN3 mRNA under conditions where global translation is relatively unaffected (Danaie et al., 1999). In mammals, aberrantly high expression of translation factors is correlated with a number of types of cancer, possibly because it increases the rate of translation of normally inefficiently translated mRNAs that encode potential oncogenic proteins (Dua et al., 2001; Watkins and Norbury, 2002).

The kinetic properties of initiation factor interactions are likely to be crucial in determining the nature of translational control. For example, translation of the GCN4 mRNA requires the same eIF2–ribosome interaction as any other cellular mRNA; yet a change in the rate of complex formation has a much enhanced effect on this particular transcript (Hinnebusch, 1997). Our understanding of the kinetic principles underlying such control phenomena is still very limited, as details of the relevant concentrations and kinetic constants involved are scarce.

The present study was therefore designed to generate information on intracellular concentrations of eIFs in the model organism S. cerevisiae and, for a subset of factors, to investigate in what way the concentrations of specific factors relate to the achievement of optimal translation rates.


Quantification of intracellular initiation factor levels

Our initial aim was to establish a method for the quantitative extraction of proteins from yeast cells, and to determine the amounts of the respective translation initiation factors in the resulting extracts. Previous estimates of initiation factor levels have frequently relied on two-dimensional gel electrophoresis and metabolic labelling with [35S]-methionine as the basis for quantification (Duncan and Hershey, 1983; Futcher et al., 1999). The main problem with this method is to assess how many gel spots represent the total cellular population of a given protein. Post-translational modifications or loss of a few charged amino acids from the ends of the polypeptide chain can dislocate subpopulations of the protein along both the pH and the size axes. This is problematic as many initiation factors occur in phosphorylated forms. For example, yeast eIF4E and p20 each contain multiple phosphorylation sites and are known to exist in heterogeneous populations in the cell (Zanchin and McCarthy, 1995).

In a more straightforward, if highly work-intensive, alternative, we opted for a Western blotting approach employing standard one-dimensional SDS-PAGE. This circumvents the problems described above, because protein modifications either do not affect the location of the protein of interest, or are readily visible as multiple bands. The realization of this strategy required that we purify eight recombinant eIF proteins or complexes and raise polyclonal antibodies against them in rabbits. Three other antibody preparations were obtained from other laboratories. We identified two main requirements for an extraction procedure suitable for protein quantification. First, the protein content of the extracts should as closely as possible reflect that of the growing cell, and the proteome should remain as stable as possible during cell harvest and lysis. Second, cells should be extracted quantitatively, thus enabling us to relate the estimated protein contents of the extract reliably to the true constitution of living cells. Of several protocols tested, we found that a single passage through a French cell press in cold denaturing buffer gave the best results in terms of rapidity and limitation of proteolysis. However, extraction in this way was not quantitative: under the conditions described in the Experimental procedures section; only approximately 50% of the cells were lysed by a single pressure cell passage.

In order to quantify reliably the number of lysed cells that gave rise to individual extracts, we performed an initial calibration experiment. A suspension containing 109 cells ml−1 was repeatedly passed through the pressure cell. After each passage, an aliquot of the suspension was taken, cleared by centrifugation and the total amount of protein determined using the BioRad protein assay. The amount of protein was then plotted against the number of extraction cycles, yielding an asymptotic curve that converged on the maximum amount of protein extractable with this method. This value was determined via curve fitting as 4.95 mg/109 cells. Additionally, analysis of the extracts on Coomassie brilliant blue-stained gels showed that the increase in protein after each passage was evenly distributed over all visible bands. The extract yields obtained from single pressure cell passages could therefore be correlated with the number of lysed yeast cells by measuring their total protein content and relating it to the value of 4.95 mg per 109 cells.

Standards containing known amounts of the respective initiation factors were calibrated by staining denaturing polyacrylamide gels with the fluorescent dye SyproRed (Molecular Probes). The fluorescence intensities of bands containing known amounts of BSA were then compared with the fluorescence of bands resulting from the electrophoresis of eIF-preparations (Fig. 1A). This method was chosen as it can selectively determine the protein contained in bands of interest, but excludes impurities or breakdown products. In contrast to classical protein quantification methods like the Bradford or Lowry assays, staining with SyproRed is highly proportional to protein mass, and largely independent of amino acid composition (Steinberg et al., 1996).

Figure 1.

Quantification of initiation factors in S. cerevisiae extracts.

A. Standards of recombinant eIFs were produced by fluorescent staining of SDS-PAGE gels and comparing the amount of protein present in the band corresponding to each full-length protein to bands containing known amounts of BSA.

B. Western blots were probed with FITC-labelled antibodies. Fluorescence associated with bands containing known amounts of eIF was then compared to fluorescence associated with bands produced from yeast extracts.

C. Representative blots obtained for the four members of the eIF4 group of initiation factors in yeast.

Figure 1 illustrates the overall procedure applied to the different proteins, and shows sample results for the eIF4 group of factors. Figure 2 shows the final results for all factors for which we could obtain unequivocal results. The mRNA and ribosome levels were those reported previously for S. cerevisiae strains growing under conditions similar to ours ( Holstege et al., 1998 ; Warner, 1999 ). It should be noted that a successful analysis could not be performed for all of the factors we tested. In the cases of Pab1 and the eIF3 subunit Nip1 (eIF3c according to the nomenclature of Browning et al., 2001 ), we found that each of the protein populations was distributed over a number of partially degraded bands and could not therefore be accurately quantified. We do not know whether this degradation represents the normal state of these proteins in vivo, or whether they were degraded during our extraction procedure. In the case of eIF2Bɛ, the signal generated by the FITC-labelled antibodies (see Experimental procedures se ction) was too weak to be detected. The amount of this factor was therefore judged to be less than 30000 molecules per cell by comparing the minimum amount of recombinant eIF2Bɛ that produced a detectable band.

Figure 2.

Intracellular initiation factor levels in Saccharomyces cerevisiae strain 1773.

A. Estimates of eIF amounts per cell. Error bars and ranges indicate the confidence levels determined for extracts prepared from three independently grown cultures, each blotted twice. Values for mRNAs per cell are taken from Holstege et al., 1998, for ribosomes per cell from Warner, 1999, and were determined with yeast strains other than 1773.

B. Abundance of yeast eIFs in order of concentration, as estimated assuming a cell volume of 27µm 3 .

In the case of eIF4G, the blot shown in Fig. 1C and the data presented in Fig. 2 were generated using antibodies that detected only the eIF4G1 isoform of this factor Repetition of the experiment with an antibody that could recognize both isoforms yielded results within the standard deviation for the first antibody, leading us to conclude that eIF4G1 represents by far the most abundant form of cellular eIF4G. This is consistent with an mRNA ratio of 5 : 1 (TIF4631:TIF4632) (Holstege et al., 1998). It is interesting to note that the Codon Adaptation Indices (from the Saccharomyces Genome Database, http:genome-http:www.stanford.eduSaccharomyces/) for eIF4G1 and eIF4G2 (Table 1) are consistent with a somewhat higher level of expression of the TIF4631 ORF (Futcher et al., 1999).

Table 1. . Summary of information relevant to eIF gene expression.
Factor Prot./cellmRNA/cellaCodon adapt.
5′UTR length Pred. sec. struct.
of 5′ UTR (kcal/mol)c
Other 5′
UTR features
  1. a .  From Holstege et al. (1998 ).

  2. b .  From SGD ( http:genome-http:www.stanford.eduSaccharomyces /).

  3. c .  Predicted using RNAfold ( Mathews et al., 1999 ).

eIF1250 00012.50.2390.24320 (Yoon and
Donahue, 1992
 − 1.5 
eIF1A 50 000 7.10.4040.357Not publ.− 17.8
(80 nt)
eIF2α180 000 3.70.3710.347Not publ. − 7.1
(80 nt)
eIF2β180 000 5.80.2850.31364 (Donahue
et al., 1988
 − 6.3 
eIF2Bα 30 000 2.40.1400.029Not publ. − 7.5
(80 nt)
eIF3g100 000 8.30.2490.298Not publ.− 13.7
(80 nt)
eIF4A800 00021.4/17.00.753/0.7510.772/
Not publ. − 1.8/−9.1 
eIF4B155 000 5.20.3510.443Not publ. − 8.5
(80 nt)
eIF4E340 00015.60.3870.38084 (Altmann
et al., 1987
− 15.1 
eIF4G1 17 500 3.40.2510.324295 (Goyer
et al., 1993
− 56.96 uORFs
eIF4G2ND 0.70.1770.142∼300 (Goyer
et al., 1993
− 68.4uORFs
p20350 000 7.30.3560.377Not publ. − 8.1
(80 nt)

The mRNA levels for all factors represented in Fig. 2 have been published previously (Holstege et al., 1998). A comparison of the respective values with our estimated protein levels is represented in Fig. 3. The correlation between mRNA and protein levels for initiation factors appears to be within the limits determined in an earlier study for a set of randomly selected yeast proteins, albeit with a considerably higher average protein:mRNA ratio than reported there (Futcher et al., 1999). This indicates that translation factor mRNAs are, as a group, relatively efficiently translated.

Figure 3.

Comparison with data from microarray analysis. Correlation between estimates for protein molecules per cell (this study) and mRNA molecules per cell ( Holstege et al., 1998 ) for yeast eIFs.

A wide range of expression levels for the eIF genes

Our estimates of the eIF protein concentrations indicate that their respective copy numbers per cell are spread over approximately a 30-fold range (Fig. 2). Distinct efficiencies of gene expression will make a significant con­tribution to these different values. We have summarized information that can provide clues as to the operation of post-transcriptional mechanisms that might generate differential rates of expression of the eIF genes (Table 1). This summary highlights some apparent correlations between the respective data sets. For example, there is a limited correlation evident between the amount of protein generated per mRNA molecule and the codon adaptation index. On the other hand, some of the eIFs with particularly low protein:mRNA ratios would be predicted to have especially stable internal secondary structure within their 5′ UTRs (eIF4G1 and eIF4G2). This would be an expected correlation if the translation of these particular mRNAs was inhibited at the initiation (scanning) step. The predicted stabilities derived here for complete 5′ UTRs are useful only as a rough guide, as the density of stable base-pairing within localized regions is likely to be the direct determinant of scanning rates (McCarthy, 1998). Unfortunately, the 5′ ends for most of the 5′ UTRs have not been experimentally determined, which means that many of the theoretical structural analyses have had to be based on an assumed length of 80 nucleotides. Finally, we note that uORFs may be responsible for limiting the translation of certain of the eIF mRNAs. The presence of uORFs has been experimentally verified in the case of what seems to be the relatively poorly translated eIF4G1 mRNA. Moreover, examination of the chromosomal sequence upstream of the eIF1A gene reveals that, depending on the actual position of the 5′ end, the encoded mRNA could have up to four uORFs in its 5′ UTR. Finally, we conclude that certain known, or predicted, structural features of the respective eIF mRNAs may help provide explanations for different rates of (post-transcriptional) gene expression. Further work is needed to clarify these potential structure-function relationships at the mRNA level.

eIF4E levels and translational efficiency

As illustrated in Fig. 2B, the protein:ribosome ratios for most of the initiation factors cluster around the value 1 : 1. Apart from eIF2B the greatest deviations from this ratio occur within the eIF4 group, which contains the factors with the lowest and the highest abundance respectively. In particular, we were interested in understanding why yeast maintains a marked excess of eIF4E over its binding partner eIF4G, as the functional unit of these two proteins for the purpose of translation is assumed to be a stably bound 1 : 1 complex (Ptushkina et al., 1998). This question is especially important because it is frequently suggested in the literature that eIF4E is a ‘limiting factor’ in translation.

In order to address the question whether the eIF4E abundance we observed in this study is strictly required for maximum translation, we employed the tet-promoter substitution technique described elsewhere (Belli et al., 1998). The 198 nucleotides immediately upstream of the chromosomal CDC33 start codon were exchanged for a cassette that places the production of eIF4E under the control of a tetO7 promoter (Fig. 4A), yielding strain YTH3. In the unrepressed state, this strain maintains eIF4E at approximately 30% of the wild-type level (Fig. 4B). This can be gradually reduced by addition of 0.01–2.0 mg l−1 doxycycline to the growth medium (compare also Fig. 6B). In contrast, in the parent strain BY4741, eIF4E levels, growth rates and translation are independent of the doxycycline concentration (Fig. 4B). Levels of eIF4G and of a protein not related to translation (Adh1) remain largely unaffected in both strains.

Figure 4.

A yeast strain in which eIF4E production is regulatable.

A. Replacement of the CDC33 promoter. The 198 bases upstream of the CDC33 start codon were replaced with a cassette containing a selectable marker and the tetO7 promoter element ( Belli et al., 1998 ).

B. eIF4E levels in the resulting strain YTH3 are reduced to 30% of wild-type levels, and can be further reduced by addition of doxycycline to the growth medium, whereas the levels of Adh1 and of eIF4E∏s binding partner eIF4G remain unaffected. In the YTH3 parent strain BY4741, eIF4E levels do not respond to doxycycline addition.

Figure 6.

The relationship between p20 and eIF4E levels.

A. Co-IP experiments reveal that eIF4E in logarithmically growing yeast is associated with both eIF4G and p20. Quantification of the precipitated material using recombinant standard proteins gives a ratio of eIF4E-bound p20 to eIF4E-bound eIF4G of 9 : 1.

B. p20 does not detectably associate with poly(A) RNA: when cell extracts prepared from cycloheximide-treated cells are incubated with oligo(dT) cellulose, both eIF4G and eIF4E can be eluted from the resin. In contrast, no p20 is detectable in the eluate. For comparison, 5 µl of total cell extract was loaded onto the gel (‘Input’) next to an aliquot of the eluted fraction corresponding to 50 µl original volume.

C. Western blots of cell extracts derived from equal amounts of cells show that in the strains containing reduced levels of eIF4E, p20 content is reduced proportionately.

D. While the p20 content of cells depends on the level of eIF4E, the reverse is not true, as a CAF20 deletion strain and its p20-containing parent contain comparable levels of eIF4E.

Figure 5 shows representative results of experiments characterizing the effect of reduced eIF4E levels on yeast. YTH3 cells grown in the absence of doxycycline show a significant increase in the average cell volume (panel A), which becomes more severe in the presence of the antibiotic. Despite this apparent change in cell morphology, YTH3 is still able to maintain average growth rates that are 80–90% of that supported by wild-type eIF4E levels. Further reduction of eIF4E abundance by addition of doxycycline to the growth medium then results in a large increase in doubling time ( Fig. 5B ). Translational activity, measured via [ 35 S]-methionine incorporation, shows a similar pattern ( Fig. 5C ). Whereas YTH3 showed average incorporation rates of 70–90% compared to its isogenic parent BY4741 in the absence of doxycycline, repression of tet -promoter activity through the addition of the anti­biotic resulted in greatly reduced translational activity. Inspection of polysome profiles generated at different eIF4E levels ( Fig. 5D ) confirms the trend observed in the previous experiments, although here YTH3 shows a distinct profile even when the promoter is fully derepressed. In conclusion, the minimal sensitivity of cellular translation and growth levels to reductions in the abundance of eIF4E is not consistent with the operation of a special ‘limiting’ influence by this factor.

Figure 5.

Effects of reductions in eIF4E levels on S. cerevisiae .

A. Pictures taken at equal magnifications show an increase in cell volume in YTH3.

B. eIF4E production under basic tetO7 promoter activity supports growth at near wild-type levels, whereas the presence of doxycycline markedly reduces the growth rate.

C. [ 35 S]-methionine incorporation rates show a similar pattern as growth rates, and are only strongly reduced in the presence of doxycycline.

D. Polysomal profiles show progressive restructuring in response to decreasing eIF4E content of cells.

The relationship between the levels of p20 and eIF4E

Saccharomyces cerevisiae produces a small (18.2 kDa) protein, the product of the CAF20 gene termed p20, which is regarded as a potential analogue of the human 4E-BP translational repressors ( Altmann et al., 1997 ; Ptushkina et al., 1998 ). This protein can bind to eIF4E and thereby exclude eIF4G, thus rendering eIF4E inactive for the purposes of translation initiation. The relationship between p20 and eIF4E is therefore relevant to the functional significance of the observed level of eIF4E in the cell. We accordingly investigated the possibility that yeast cells have a pool of eIF4E that is maintained in a complex with p20 and is therefore not capable of binding to eIF4G. Coupled with that possibility is the question whether p20 stabilizes the apparently unbound pool of eIF4E.

We found indeed that p20, as well as eIF4G, can be immunoprecipitated with anti-eIF4E antibodies from extracts prepared from logarithmically growing cells (Fig. 6A). In order to minimise potential displacement of eIF4E-bound proteins by antibodies that share the same binding site as eIF4G and p20, we kept the contact between cell extract and preimmobilized antibodies very short (10 min, see Experimental procedures). According to the kinetic constants that we previously measured for the respective interactions (Ptushkina et al., 1998), this should ensure maintenance of more than 80% of the initially formed complexes by the end of the incubation period. Quantification of the co-precipitated proteins, using recombinant protein standards as described above, revealed a p20:eIF4G ratio of 9 : 1, similar to the estimated ratio of these proteins in the cell.

The presence of a large pool of eIF4E that is bound to a potential translational inhibitor raises the question whether the 5′ cap structures of translated mRNAs exist in an equilibrium of p20- and eIF4G-associated states, or whether the translationally active eIF4E:eIF4G complexes can exclude the inactive eIF4E:p20 complexes from the caps. Initial fractionation experiments showed that p20 is not present in large amounts in ribosomal pellets after ultracentrifugation of cell extracts. The interpretation of these experiments is difficult as non-polysome associated proteins always occur as contaminants in the ribosomal pellets. However, p20 appears not to be polysome-associated because its distribution is comparable to that of Adh1, with 90% of the protein in the post-ribosomal supernatant. In contrast, 50–70% of eIF4G occurs in the ribosomal pellets (data not shown). In order to obtain more reliable data on the mRNA-binding status of the different eIF4E-containing complexes, we isolated total poly(A)RNA via passage of cell-extracts over oligo(dT) cellulose. In eluates derived from these experiments, eIF4G and eIF4E can be clearly detected (Fig. 6B). In contrast, no p20 could be detected bound to the resin, despite the fact that our anti-p20 antibodies are two to three times more sensitive than the anti-eIF4E antibodies (data not shown). It appears therefore that the eIF4E:p20 complexes are not bound to mRNA 5′-caps under conditions of logarithmic growth. The most likely mechanism of exclusion arises from the different co-operativity effects exerted by p20 and eIF4G: we had initially shown that p20 arrests eIF4E somewhat less tightly on cap-analogues than eIF4G (Ptushkina et al., 1998); in addition, the binding of Pab1 to eIF4G was shown to have a further stabilizing influence (von der Haar et al., 2000). It was therefore of particular interest to find that Pab1 is also present in detectable amounts in the poly(A)+ isolate (data not shown).

Interestingly, we found that alterations of the eIF4E levels in YTH3 are closely followed by changes in p20 abundance (Fig. 6C). However, the reverse is not the case: a caf20 deletion strain contains eIF4E levels that are indistinguishable from those of its p20-producing counterpart (Fig. 6D). In conclusion, we believe that p20 is not involved in maintaining a high eIF4E:eIF4G ratio; we assume that the co-regulation with eIF4E simply stems from rapid degradation of non-eIF4E-bound p20 molecules, and therefore that eIF4E influences p20 stability and therefore abundance in vivo. Further observations support this hypothesis. First, p20 levels cannot be detectably increased even if an additional copy of the gene coupled to a strong promoter is introduced into yeast. Second, in vitro the protein is extremely difficult to handle even in pure form, and is more stable in solution in the presence of eIF4E (T. von der Haar, Geffers and J. E. G. McCarthy, unpublished data).

Finally, it should be noted that there has been uncertainty about the phenotype associated with deletion of caf20 (Altmann et al., 1997; de la Cruz et al., 1997), with only one group describing a very moderate increase in growth rate (Altmann et al., 1997). In our hands, a strain lacking p20 did not show significantly altered generation times at a variety of temperatures or under osmotic or oxidative stress conditions, nor did it respond measurably differently during adaptation to, or recovery from, these stresses (data not shown). We suspect therefore, that p20 may bind to the pool of eIF4E that is not engaged in a translationally active complex with eIF4G. This pool may even not be involved directly in translation, but may rather have a quite distinct function, perhaps linked to its nuclear localization (see Discussion). In a separate study, we have found that the overproduction of both eIF4E and p20 in yeast can affect mRNA stability (Vilela et al., 2000; Velasco et al., 2002).


Factor quantification

The scientific literature contains a number of observations that can be used for the purpose of cross-referencing and validating the results obtained in this study. In particular, eIF4A has previously been described as the most abundant initiation factor (Neff and Sachs, 1999), and a roughly equimolar occurrence of eIF4E and p20 with the ribosomes was observed, but a lower abundance of eIF4G (Altmann et al., 1997). Correlation of our protein values with published mRNA copy numbers for initiation factors generally supports the relative ratios between eIFs that we found. The factor with the highest protein copy number also has the highest mRNA abundance, and the converse relationship exists for the two least abundant eIFs (Fig. 3). The ratios for the other factors do not correlate as well, but are within published ranges for mRNA:protein ratios (Futcher et al., 1999).

Although yeast cell volume can change considerably, for example as a function of carbon source (Tyson et al., 1979), analyses under conditions very similar to ours have consistently found volumes of 35–40 µm3 (Tyson et al., 1979; Russo et al., 2000; 2001). Of this total volume, 25% has been estimated to be solid material (Dombek and Ingram, 1986). Accordingly, the intracellular volume available to be occupied by macromolecules corresponds to 25–30 µm3 per cell, assuming an initial value of 36 ± 3 µm3 (Tyson et al., 1979). With the content of extractable protein estimated by us to be 5 mg per 109cells, this would result in total intracellular protein concentrations of 170–200 mg ml−1, which compares to published values of 200–300 mg ml−1 cytoplasmic protein for E. coli (Zimmerman and Trach, 1991). These calculations neglect the fact that yeast as a eukaryote is organized into compartments (including vacuoles), which may contain significantly different densities of macromolecules. They are intended to demonstrate that we are unlikely to have made any gross errors in our calculations of cell numbers and protein concentrations.

The results presented here were obtained through application of a single method, and may therefore suffer from inherent inaccuracies. Also, the concentrations of specific proteins are likely to be at least as variable as is cell size between different populations. Taking these considerations into account, the minimal interpretation of the data presented in Fig. 2 is to divide yeast translation initiation factors into three groups: those that are significantly more abundant than ribosomes (eIF4A), those that are less abundant (eIF1A, eIF2Bα and eIF4G), and those that are roughly comparable to ribosome levels (i.e. within a range covering 0.5–2.0 × the level of ribosomes: eIF1, 2α, 2β, 3c, 4B, 4E and p20). Even when interpreting our findings in such a broad way, they have interesting implications for interactions that occur between eIFs.

Leaving aside the factors involved in cap-complex formation (the group 4 factors and p20), the groups identified above by means of abundance relate in a fully explicable way to the structural and non-structural roles of the respective factors during formation of the 43S complex. All factors occurring at levels comparable to the ribosomes have been previously identified as integral parts of the 43S complex (Asano et al., 2000). In contrast, eIF2B is well characterized as an enzyme that regenerates the eIF2:GDP complex produced after each initiation event (reviewed in Webb and Proud, 1997). The notion that this factor acts as an enzyme is consistent with the requirement for a lower abundance compared to its proposed substrate. A similar situation may be true for eIF1A, which was shown to catalyze the eIF2-dependent inline image transfer to the small ribosomal subunit (Chaudhuri et al., 1997). Alternatively, this factor may constitute a limiting structural component of 43S complexes.

Observations made after glucose withdrawal from logarithmically growing cells show that yeast can progress from the absence of translation to normal polysomal profiles within a very short time (Ashe et al., 2000). The comparable ratios between ribosomal subunits and the other 43S complex components are consistent with the ability to rapidly commit the entire 40S pool to translation. The observed equimolarity would allow most or all of the 40S subunits to simultaneously exist as 43S complexes, omitting the need to wait for a first round of ribosomes to finish initiation, eIF2 to be regenerated, and a second round of 43S complexes to be formed.

In contrast to the proteins discussed above, the observed ratios between the eIF4 factors are less straightforward to explain. In the case of eIF4A, the very high abundance is consistent with observations that this protein has relatively weak helicase activity compared to other members of the DEAD-box family (de la Cruz et al., 1999), and that it was proposed to function in translation in a manner that involves cycles of transient binding to a more stably bound eIF4E:eIF4G complex (Yoder-Hill et al., 1993; Pause et al., 1994). Thus, the large amount of intracellular eIF4A would be required to enable delivery of a sufficient level of helicase-activity to mRNA 5′ ends, a notion that is supported by the observation that addition of eIF4A to yeast cell-free extracts increases translational activity despite the high initial levels of this protein (Neff and Sachs, 1999). Interestingly, eIF4A helicase-activity is stimulated by eIF4B in vitro (Rozen et al., 1990); but the eIF4A:eIF4B ratios reported here indicate that this stimulation can only apply to a subset of eIF4A molecules at any given time. It is known that eIF4A can function in translation independently of eIF4B, as strains unable to produce the latter protein are viable (Coppolecchia et al., 1993). Taken together, these findings suggest that eIF4A-related RNA helicase activity may be delivered in two different modi in vivo, partially by eIF4A alone, and partially by eIF4B-stimulated eIF4A.

Assuming again a total intracellular volume available to proteins of 25–30 µm3, the concentration of ribosomes in yeast would be as high as 11–13 µM; that of eIF4A would be 50–60 µM and the concentration of eIF4G or eIF2B would still be in the µM range (Fig. 2B). Binding partners interacting with equilibrium constants in the low to mid nM range (kDa = 10−9–10−7), such as eIF4E:eIF4G (Ptushkina et al., 1998), or eIF3:eIF1 (von der Haar et al., 2002), would therefore be very close to existing fully in the bound state at any given time provided that these interactions are not regulated by other proteins.

More interesting still is the consequence of such high concentrations for weaker interactions. Transient interactions, and the ability to change readily from a bound to an unbound state, are an inherent requirement of a model of translation initiation in which one ribosome after another is recruited to an mRNA by the action of initiation factors. Affinities in the range of 1–10 µM best fit this specification for the assumed factor concentrations, whereas stronger affinities would make it less likely for a protein to spend significant time in the unbound state. Conversely, weaker interactions would reduce the likelihood of efficient recognition of the binding partner. However, it is important to note that according to thermodynamic theory such calculations have to be based on activities rather than concentrations. An important next step in assessing the impact specific translation initiation factor levels have on translational regulation will therefore be the experimental determination of coefficients that relate concentration to activity in the densely crowded intracellular environment.

In conclusion, the eIF quantification performed in this study has generated information of key significance to our understanding of translation in a eukaryotic model organism. The numbers obtained from this work con­-stitute an important guide for the next major step in research on the yeast translation system, which will be the full reconstitution and quantitative analysis of its components.

Intracellular eIF4E concentrations and translation

As has been noted above, the levels of eIF4E, eIF4G and p20 reported here cannot easily be related to our current models of translation initiation. According to the commonly accepted model, supported by an extensive body of evidence, eIF4E functions in a 1 : 1 complex with eIF4G to build a molecular bridge between the mRNA 5′ cap and the 43S complex. If this is the case, why does yeast maintain an almost 10-fold excess of eIF4E over eIF4G? Theoretically, an excess of eIF4E might be important for translation initiation, e.g. in order to enforce high rates of complex formation; alternatively, eIF4E may be required independently of eIF4G in other areas of mRNA metabolism (for a review of roles of eIF4E outside translation see Strudwick and Borden, 2002). Our experiments investigating the effects of limitation of intracellular eIF4E levels by promoter substitution very clearly show that the bulk of ongoing translation under logarithmic growth conditions does not require most of the available eIF4E. This finding is consistent with previous studies showing that cap-binding activity of eIF4E can be considerably weakened by mutagenesis without severely affecting cell viability (Vasilescu et al., 1996), which had already indicated that eIF4E does not exert particularly strong rate limitations on translation in yeast cells. Moreover Lang et al. (1994) showed that changes in the in vivo abundance of eIF4E over a broad range have little effect on cell viability. Here, we quantify for the first time the extent to which a reduction in eIF4E levels can be tolerated, in that more than two thirds of eIF4E can be lost without affecting either growth rate or translational activity.

In view of the assumed analogy between the mammalian 4E-BPs and yeast p20, it was surprising to find that p20 is associated with eIF4E during logarithmic growth. The suggestion that p20 has the same function of translational repressor as the 4E-BPs have in mammals was based mainly on the biochemical similarities between the two proteins: both are small phosphoproteins, they can bind eIF4E via motifs that are very similar to each other and motifs present in eIF4G, and when doing so, they displace eIF4G, thereby rendering eIF4E inactive for translation (Altmann et al., 1989; 1997; Zanchin and McCarthy, 1995; Ptushkina et al., 1998). However, in contrast to the 4E-BPs, it has never been shown for p20 that association with eIF4E depends on the translational status of the cell, or on the phosphorylation status of p20.

The association between p20 and eIF4E under conditions of high translational activity suggests that its main function may not be as translational inhibitor. Analysis of CAF20 deletion strains, and of strains containing reduced eIF4E levels, shows that the presence of eIF4E:p20 complexes is not required for the control of translation under standard growth conditions, nor for any other aspect of cellular function investigated so far. Because we could not detect association of p20-bound eIF4E with poly(A) mRNAs in wild-type yeast during logarithmic growth, we conclude that logarithmically growing yeast maintains eIF4E in two functional pools: one, bound to eIF4G, is necessary for translation of all capped mRNAs. The other, bound to p20, is not required for translation at least of the majority of mRNAs. Although the eIF4E normally bound in this pool is not essential for bulk translation and normal growth rates, it or part of it is essential for the maintenance of normal cellular morphology: this may be because it is necessary for translation of particular mRNAs with higher requirements on eIF4E activity, such as the CLN3 transcript (Danaie et al., 1999), or because it has roles outside of translation.

Although the pool of non-eIF4G-bound eIF4E is bound to p20 in wild-type yeast cells, it can exist independently of the latter protein if this is absent from the cells. Therefore, in contrast to the eIF4E contained in the pool of eIF4E:p20 complexes, p20 itself appears not to be involved in cellular fitness under any of the conditions tested. So far, the only clear phenotype observed when levels of p20 and eIF4E are co-ordinately increased is a modulation of half-lives of mRNAs in vivo, presumably stemming from competition with the mRNA degradation machinery (Velasco et al., 2002). It would therefore be interesting to test whether p20 has a function under conditions where RNAs have to be protected from the degradation machinery in the absence of translation. An example for such a situation is sporulation, during which the CAF20 gene was shown to be transcriptionally regulated (Chu et al., 1998).

A last point of discussion is whether the situation described here for the cap-binding complex of budding yeast is specific to this organism or applies to eukaryotes in general. p20 appears to be significantly different from the 4E-BPs in that it is associated with eIF4E even under conditions of the highest translational activity, in complexes that are not bound to polyadenylated mRNAs. A role for this protein, whatever it may eventually turn out to be, is therefore possibly specific to yeast.

In terms of the excess of eIF4E over eIF4G, we cannot yet determine whether this situation is particular to yeast or occurs also in other cells. To our knowledge, the only other study that has investigated intracellular levels of eIF4E and eIF4G concerned HeLa cells, where the situation appears reversed compared to yeast with a twofold excess of eIF4G over eIF4E (Duncan et al., 1987). However, changes in the levels of both eIF4E and eIF4G have been reported for many cancers (Dua et al., 2001; Watkins and Norbury, 2002), so that the ratios for these factor may be different in non-transformed cell lines. Indeed, it has previously been questioned whether a general description of eIF4E as the rate limiting factor in translation is true even for mammalian cells (Rau et al., 1996, also discussed in McCarthy, 1998).

Experimental procedures

Yeast strains

Yeast strains used in this study are summarized in Table 2.

Table 2. . Yeast strains used in this study.
1773MATaarg9 Esteban and
Wickner (1988
BY4741MATahis3Δ1 leu2Δ0
met15Δ0 ura3Δ0
BY4742MATαhis3Δ1 leu2Δ0
lys2Δ0 ura3Δ0
MATαhis3Δ1 leu2Δ0 lys2Δ0
ura3Δ0 caf20::kanMX4
YTH3MATahis3Δ1 leu2Δ0
met15Δ0 ura3Δ0
This study
CDC33(− 198,-1)::KanMX4-

Preparation and standardization of protein extracts

The yeast extracts for the determination of values presented in Fig. 2 were prepared as follows: S. cerevisiae strain 1773 was grown in a shaking water bath at 30°C to mid-log phase (5 × 106−2 × 107 cells ml−1), at which point all cultures used in this study had a doubling time of 80 ± 3 min. Cells were harvested at 4°C in pre-cooled containers, and the pellet resuspended to a density of 109 cells ml−1 in ice-cold extraction buffer (20 mM Hepes pH 7.5, 6 M urea, 100 mM KCl, 10 mM EDTA, 5 mM β-mercaptoethanol, 1 mM PMSF). The suspension was passed through a precooled French pressure cell press at 16–18000 p.s.i. Extracts for blots represented in Fig. 4 and onwards were generated by freeze-thawing and boiling a detergent-containing cell-slurry, and subsequent application of the non-centrifuged slurry to SDS-PAGE gels as described in (Langlands and Prochownik, 1997).

Preparation of recombinant eIF standards

Recombinant eIF4E and histidine-tagged eIF4G were prepared as described previously (von der Haar et al., 2000). Histidine-tagged eIF2 was prepared from yeast strain GP3511 as described (Pavitt et al., 1998). Histidine-tagged eIF3 was purified from yeast strain LPY87 as described (Phan et al., 1998). Histidine-tagged eIF4B was expressed in Sf9 insect cells and purified using standard immobilized metal chelate chromatography followed by chromatography on a Heparin column. Untagged eIF4A was expressed in E. coli and purified using ion exchange chromatography followed by size exclusion chromatography. eIFs 1, 1A, 5, and p20 were expressed in Histidine tagged form in E. coli, and purified using a combination of immobilized metal chelate chromatography and ion exchange chromatography.

Quantitative Western blotting

Antibodies against initiation factors were generated by immunization of rabbits with the purified factors (Abcam, Cambridge, UK), except for antibodies specific to eIF4G1, which were generated by immunization with the histidine-tagged N-terminal half (amino acids 1–513) of yeast eIF4G1 purified from E. coli (von der Haar et al., 2000). Antibodies against eIF4E have been described earlier (Lang et al., 1994). For Western blotting, standardized amounts of recombinant proteins were applied to SDS-PAGE gels next to proteins extracts. After separation, the proteins were transferred onto PVDF membranes using a submersed blotting apparatus (BioRad). The membranes were then blocked in 1% gelatine in TBS (50 mM Tris-HCl pH 7.8, 150 mM NaCl), incubated overnight with primary antibodies in their respective optimal concentrations mixed with FITC-coupled anti-rabbit IgG (Sigma) at a dilution of 1 : 2000, and washed.

Quantification of band-associated fluorescence in gels and Western blots

SDS-PAGE gels were stained with SyproRed (Molecular Probes) according to the manufacturer's instructions. They were then placed directly on the scanning surface of a Typhoon Imager (Molecular Dynamics), and scanned using green laser excitation and an emission filter of 610 nm. Western blots were dried and scanned using green laser excitation and an emission filter of 526 nm. Fluorescence values associated with bands in the scanned images were determined using ImageQuant software (v 5.1, Molecular Dynamics).

Construction and characterization of strain YTH3

This strain was constructed by transformation of S. cerevisiae BY4741 with a PCR product generated from plasmid pCM225, using the primers 5′-TTTTTTTCGTCGTCAAT AGAGTTTAATGCAATACCTGATACAGCTGAAGCTTCGTAC GC-3′ and 5′-CGTTTTCTTCAAACTTCTTGCTAACTTCTT CAACGGACATATAGGCCACTAGTGGATCTG-3′. This resulted in a strain in which the 198 nucleotides upstream of the CDC33-AUG codon are replaced by the tetO7-promoter cassette described in (Belli et al., 1998). In order to obtain logarithmic growth cultures at different intracellular eIF4E concentrations, aliquots of medium containing various amounts of doxycycline (Clontech) were inoculated from stationary phase cultures and grown overnight. These cultures were then used to inoculate a further aliquot of medium to the desired initial cell density. Cell densities were determined using an improved Neubauer chamber. Polysomal gradients were prepared from cultures grown to OD = 0.7 as previously published (Sagliocco et al., 1993). For the isolation of poly(A)-RNA bound proteins, cell extracts were prepared from cycloheximide-treated cells as for polysomal gradients, except that a salt concentration of 300 mM was used. Extract from 100 ml culture grown to OD = 0.8 was incubated with 50 µl equilibrated oligo(dT) cellulose (Sigma O3131 at 4°C for 1 h. The resin was then washed four times with 1 ml of incubation buffer containing 0.05% Tween 20, and bound protein eluted by heating in one-tenth of the original volume of SDS-PAGE sample buffer).

[35S]-methionine incorporation was monitored as described (Sachs and Deardorff, 1992).

Co-immunoprecipitation experiments

These were performed as follows: 30 µl of protein A agarose (Sigma) were incubated with 100 µl of rabbit antiserum. The resin was washed once with 1 ml of buffer (20 mM Hepes pH 7.5, 100 mM KCl, 5 mM MgCl2, 5 mM β-mercaptoethanol, 1 mM PMSF) and incubated for 10 min with 200 µl of a cleared yeast extract, prepared in the same buffer from logarithmically growing cells via passage through a pressure cell. The resin was then washed with 3 × 1 ml of buffer, and bound proteins eluted in 50 µl of standard SDS-PAGE sample buffer by heating to 95°C for 5 min.


We are indebted to the following for supplying antibodies and proteins: Dr Graham Pavitt (UMIST, UK) for eIF2B subunits and antibodies as well as yeast strain GP3511 for the purification of eIF2; Dr Mark Ashe (UMIST, UK) for an antibody recognising both yeast eIF4G isoforms; and Dr Alan Hinnebusch (NIH, Bethesda, MD) for antibodies against eIF3 subunits and yeast strain LPY87 for the purification of eIF3. We thank the BBSRC and the Wellcome Trust for supporting this research.