Efficient incorporation of eukaryotic initiation factor 1 into the multifactor complex is critical for formation of functional ribosomal preinitiation complexes in vivo.

Eukaryotic initiation factor 1 (eIF1) is a low molecular weight factor critical for stringent AUG selection in eukaryotic translation. It is recruited to the 43 S complex in the multifactor complex (MFC) with eIF2, eIF3, and eIF5 via multiple interactions with the MFC constituents. Here we show that FLAG epitope tagging of eIF1 at either terminus abolishes its in vitro interactions with eIF5 and eIF2beta but not that with eIF3c. Nevertheless, both forms of FLAG-eIF1 fail to bind eIF3 and are incorporated into the 43 S complex inefficiently in vivo. C-terminal FLAG tagging of eIF1 is lethal; overexpression of C-terminal FLAG-eIF1 severely impedes 43 S complex formation and derepresses GCN4 translation due to limiting of eIF2.GTP.Met-tRNA(i)(Met) ternary complex binding to the ribosome. Furthermore, N-terminal FLAG-eIF1 overexpression reduces eIF2 binding to the ribosome and moderately derepresses GCN4 translation. Our results provide the first in vivo evidence that eIF1 plays an important role in promoting 43 S complex formation as a core of factor interactions. We propose that the coordinated recruitment of eIF1 to the 40 S ribosome in the MFC is critical for the production of functional 40 S preinitiation complex.

Eukaryotic initiation factor 1 (eIF1) is a low molecular weight factor critical for stringent AUG selection in eukaryotic translation. It is recruited to the 43 S complex in the multifactor complex (MFC) with eIF2, eIF3, and eIF5 via multiple interactions with the MFC constituents. Here we show that FLAG epitope tagging of eIF1 at either terminus abolishes its in vitro interactions with eIF5 and eIF2␤ but not that with eIF3c. Nevertheless, both forms of FLAG-eIF1 fail to bind eIF3 and are incorporated into the 43 S complex inefficiently in vivo. C-terminal FLAG tagging of eIF1 is lethal; overexpression of C-terminal FLAG-eIF1 severely impedes 43 S complex formation and derepresses GCN4 translation due to limiting of eIF2⅐GTP⅐ Met-tRNA i Met ternary complex binding to the ribosome. Furthermore, N-terminal FLAG-eIF1 overexpression reduces eIF2 binding to the ribosome and moderately derepresses GCN4 translation. Our results provide the first in vivo evidence that eIF1 plays an important role in promoting 43 S complex formation as a core of factor interactions. We propose that the coordinated recruitment of eIF1 to the 40 S ribosome in the MFC is critical for the production of functional 40 S preinitiation complex.
In eukaryotic translation initiation, the small ribosomal subunit (40 S ribosome) binds the eukaryotic initiation factor 2 (eIF2) 1 ⅐GTP⅐Met-tRNA i Met ternary complex (TC) to form the 43 S preinitiation complex. Subsequent joining of mRNA carried by eIF4F produces the 48 S preinitiation complex. The eIF3 stimulates recruitment of Met-tRNA i Met and mRNA to the 40 S ribosome by binding to eIF2 and the eIF4G subunit of eIF4F, either directly or indirectly (for a review, see Ref. 1). The 48 S complex searches for the first AUG codon in the mRNA with the help of low molecular weight factors, eIF1 and eIF1A, and the helicase and its cofactor, eIF4A and eIF4B, respectively. Correct base pairing between the Met-tRNA i Met anticodon and the AUG codon triggers the hydrolysis of GTP bound to eIF2; this reaction is dependent on the GTPase-activating function of the eIF5 N-terminal domain (NTD). This GTP hydrolysis then leads to dissociation of preassembled eIFs and formation of an initiation complex containing the AUG anticodon base pair in the ribosomal P site. The second GTP-binding protein eIF5B then stimulates joining of the 40 S initiation complex with the 60 S subunit to form the 80 S initiation complex. The elongation of the polypeptide chain starts from the methionine linked to the 80 S initiation complex. eIFs are highly conserved from yeast to mammals. Mammalian eIFs were purified from a high salt wash fraction (e.g. in a 500 mM KCl buffer) of ribosome-associated proteins and characterized in crude mammalian cell extracts or partially or fully reconstituted initiation systems (1)(2)(3). Despite the progress in elucidating the function of individual eIFs, the precise order of and the key element(s) critical for their assembly in vivo remain to be elucidated.
Using yeast Saccharomyces cerevisiae as a model organism, it was proposed that the C-terminal domain (CTD) of eIF5 plays a critical role in the assembly and integrity of the functional 43 and 48 S complexes (4,5). eIF5-CTD interacts concurrently with the ␤ subunit of eIF2 and the c subunit of eIF3 via a conserved motif called aromatic/acidic boxes (AA-boxes) 1 and 2 (6). Lysine-rich segments (K-boxes) in eIF2␤-NTD are responsible for its binding to the eIF5-CTD (6). These interactions plus interactions between eIF3c and eIF1 (7), between eIF3a and eIF1, and between eIF2␤ and eIF3a (8) were proposed to mediate formation of the multifactor complex (MFC) containing eIF1, eIF2, eIF3, eIF5, and Met-tRNA i Met (4). Accumulating evidence supports the model that the constituents of the MFC bind to the 40 S ribosome as a preformed unit to form the 43 S complex (5,9,10). Disruption of MFC by an AA-box 2 mutation in eIF5 leads to a defect in Met-tRNA i Met and mRNA binding to the 40 S ribosome in vitro. Evidence also suggests that this mutation impedes a step subsequent to the 43 S complex formation in vivo (5).
Among the components of the MFC, eIF1 is a small factor (12 kDa) whose importance has just begun to be appreciated. Genetic and biochemical studies indicate that eIF1 is required for discrimination of the 48 S complex against near cognate codon pairing with the tRNA i Met anticodon, to ensure initiation from AUG only (2,11,12). Besides its role in the 48 S complex, in vitro assays using the 40 S ribosome, Met-tRNA i Met , and a limited set of eIFs indicate that eIF1 stimulates eIF2 TC binding to the 40 S ribosome both in yeast (13) and mammals (14).
In this paper, we investigated mutual interactions of eIF1 with other components in the MFC and found that eIF1 binds to the other two components of MFC, eIF2␤ and eIF5, in addition to eIF3c, as was previously shown (7). These newly iden-tified eIF1 interactions, but not that with eIF3c, are significantly reduced by tagging of eIF1 with the highly charged FLAG peptide at either terminus. The FLAG-tagged forms of eIF1 reduce binding to eIF3 and are recruited to the 40 S ribosome inefficiently. C-terminal FLAG-eIF1 produces a recessive lethal phenotype, and its overexpression severely impedes 43 S complex formation. N-terminal FLAG-eIF1 overexpression also compromises 43 S complex formation and reduces eIF2 binding to the 40 S ribosome. Together these results provide firm evidence that eIF1 plays an important role in promoting 43 S complex formation in vivo.

MATERIALS AND METHODS
Plasmids and Yeast Strains-Plasmids and yeast strains used in this study are listed in Tables I and II, respectively. The 1.3-kb SphI fragment of p1128 (11) containing the chromosomal SUI1 locus was cloned into YCplac111 and YEplac181 (15), yielding YCpL-SUI1 and YEpL-SUI1, respectively. The 1.6-kb PstI-SphI fragment of p4231 carrying p GPD -FL-SUI1-t PGK was cloned into YCplac111 to generate YCpL- To introduce a NcoI site at the 5Ј-end of the SUI1 ORF, a SUI1 5Ј-untranslated region DNA was amplified with oligo-1 (5Ј-CCC GAG CTC GAA TCT ATT CTG GAC ATC CTG-3Ј) and oligo-2 (5Ј-GCG GGA  TCC ATG GGA TTT GCT TCA GCT ATA TTA ATA TAT TC-3Ј). Restriction enzyme sites are underlined. The SacI-BamHI fragment of this DNA fragment was replaced with the SacI-BamHI segment of YCpL-SUI1. The resulting plasmid, YCpSUI1⌬Nco, has a unique NcoI site following the 5Ј-untranslated region and lacks the 5Ј-half of SUI1 ORF up to the unique BamHI site. The 0.21-kb NcoI-BamHI fragment of p4231 containing the FL-SUI1 ORF 5Ј-half was subcloned into YCpL-SUI1⌬Nco to generate YCpL-FL-SUI1 encoding FL-eIF1 under the natural promoter. The 0.21-kb NcoI-BamHI fragment from PCR with oligo-3 5Ј-CCC CCA TGG ACT ACA AGG ACG ACG ATG ACA AGC TGT CCA TTG AGA ATC TGA AAT C-3Ј (to replace the second ATG of FL-eIF1 ORF with CTG; underlined) and oligo-4 5Ј-CTC CCC CAT TTC TGG ATC CTT GAC-3Ј (covering the unique BamHI site in SUI1 ORF) was cloned into YCpL-SUI1⌬Nco to generate YCpL-FL-SUI1* encoding another version of FLAG-eIF1 designated FL-eIF1*.
Co-immunoprecipitation was done essentially as described (6) with the following modification. Whole cell extracts (WCE) were prepared in buffer A, but immune complex binding and washing were done in buffer A supplemented with 0.1% Triton X-100. Polysome analysis was conducted as described previously (4,5).
To quantitate the amount of factors in the precipitated fractions, we used WCE prepared from a wild type strain (KAY146) as a standard, based on the following information. Using purified FL-eIF1 as a reference, we determined that 40 g of WCE contains 3.6 pmol of eIF1 (see Fig. 3A). Because the intracellular molar ratio of eIF1, eIF2, eIF3, and eIF5 is 1:0.77:0.65:0.82 in our WCE, 3 the same amount of WCE was calculated to contain 2.8, 1.9, and 3.0 pmol of eIF2, eIF3, and eIF5, respectively. The amounts of precipitated eIFs were determined in reference to these values. We determined individual eIF levels by anti-FLAG immunoblotting of WCE prepared from strains encoding FLAG-eIF1, -eIF2␤, -eIF3c, or -eIF5 as its sole source. 3 The values obtained are in better agreement with data from Ref. 18 than the data from Ref. 19.
All of the biochemical assays were done at least three times, and a typical result is shown.
To further examine interactions between eIF1 and eIF5 or eIF2␤, the bacterial extract containing a recombinant form of eIF1 (r-eIF1) was incubated with GST-eIF5-B6, GST-eIF2␤-N, GST-eIF3c-N, or GST alone. GST-eIF3c-N contains the N-terminal 156 amino acids of eIF3c, sufficient for eIF1 binding (4), and was used as a positive control. GST fusion proteins bound to r-eIF1 were one-step purified with glutathione and analyzed by SDS-PAGE. As shown in Fig. 1C, top panel, Coomassie staining of the eluted proteins indicates that GST-eIF3c-N (lane 6), but not GST alone (lane 4), specifically bound to a nearly stoichiometric amount of r-eIF1, whereas no protein from mock-induced extracts bound to GST-eIF3c-N (lane 5). Under these conditions, GST-eIF5-B6 and GST-eIF2␤-N bound recombinant eIF1, albeit weakly as determined by Western blotting (Fig. 1C, bottom panel, lanes 9 and 12; note that 5 times less eluted fraction was analyzed in lanes 5-7 than in lanes 8 -13). Therefore, r-eIF1 binds GST-eIF5-B6 and GST-eIF2␤-N. The treatment of the GST fusion complexes with RNase A prior to the elution step did not reduce these interactions (Fig. 1C, lanes 7, 10, and 13), ruling out the possibility that r-eIF1 was tethered to the proteins via RNA fortuitously bound to r-eIF1 or the GST fusion protein.
As mentioned above, the new interactions of eIF1 with eIF2␤ and eIF5 are weaker than the previously known interaction between eIF1 and eIF3c (Fig. 1C). However, it is these weaker interactions that contribute to cooperative formation of eIF1⅐eIF2␤⅐eIF5 subcomplex in the MFC, as shown in Fig. 1, E and F. Consistent with this model, FLAG epitope tagging of eIF1 compromises these interactions, leading to inefficient incorporation of FLAG-eIF1 into the MFC (see Fig. 2C).
Both the novel interactions of eIF1 depend on the AA-boxes of eIF5 and K-boxes of eIF2␤, because the interaction between GST-eIF5-B6 and 35 S-eIF1 was abolished by the AA-box 1 mutation, tif5-12A, or the AA-box 2 mutation, tif5-7A (Fig. 1A, lanes [3][4][5], and altering all of the lysines in the K-boxes to alanines reduced GST-eIF2␤ binding to eIF1 by 4-fold (Fig. 1B, lane 5). We also found that these interactions were salt-sensitive, since they were not detected when a buffer containing 150 mM Na ϩ was used at the washing step (7) instead of a buffer containing 75 mM KCl as employed in Fig. 1 (data not shown). We confirmed, however, that 35 S-eIF1 interacts with eIF5-B6 and eIF2␤-N in the same binding buffer (7) with 100 mM KCl; this salt concentration has been used in a variety of translation initiation assays (23). The salt sensitivity of these interactions explains our previous failure to identify these interactions.
To confirm that eIF1 can bind directly to the trimeric eIF2 complex, we affinity-purified FLAG-tagged eIF2 from yeast and allowed it to bind r-eIF1 expressed from bacteria. The Coomassie-stained gel of the purified eIF2 is shown in Fig. 1D, lane 2. The eIF2⅐r-eIF1 complex was precipitated with anti-FLAG affinity resin and then analyzed by Western blotting. As shown in Fig. 1D, second panel, r-eIF1 bound specifically to FLAG-eIF2 as detected by anti-eIF1 (lanes 4 -9). To examine whether this interaction is mediated by eIF3 or eIF5 associated with FLAG-eIF2, we analyzed the complex with antibodies against eIF2␣, eIF3g, and eIF5. As shown in Fig. 1D, bottom two panels, we found little eIF3 and eIF5 in the precipitated FLAG-eIF2 fractions (lanes 6 and 9), whereas the amounts of eIF1 and eIF2␣ in the FLAG-eIF2⅐r-eIF1 fraction was judged to be nearly stoichiometric (lane 6), when compared with their amounts in yeast WCE (lane 3). These results indicate that eIF1 can bind native eIF2 as well as its ␤ subunit.
eIF2␤-NTD Interacts Simultaneously with eIF1 and eIF5-CTD-Having observed separate interactions of eIF1 with eIF2␤-NTD and eIF5-CTD, we pondered whether these interactions occur simultaneously. If so, the formation of a trimeric eIF1⅐eIF2⅐eIF5 complex as a part of MFC would contribute to stable eIF2 TC binding to the 40 S ribosome. Thus, we tested whether the interaction between GST-eIF5-B6 and eIF1 can be enhanced by the addition of eIF2␤-NTD by a bridging mechanism. As a control, we used the eIF3c-NTD segment, since it is known to bridge the same interaction between GST-eIF5-CTD and eIF1 (4). As shown in Fig. 1E, bottom panel, the eIF2␤-N segment increases 35 S-eIF1 binding to GST-eIF5-B6 by forming a bridge, as efficiently as the eIF3c-N segment does (top panel). The addition of equivalent amounts of bovine serum albumin 3 H. He, C. R. Singh, and K. Asano, unpublished data.  Fig. 1E is not due to increasing the efficiency of the pull downs by nonspecific mechanisms. Thus, eIF2␤-NTD binds simultaneously to eIF1 and eIF5-CTD. Likewise, Fig. 1F shows that the His-eIF5-B6 segment can bridge GST-eIF2␤-N and 35 S-eIF1, thereby enhancing this interaction (compare lanes 2-4), indicating that eIF5-CTD binds simultaneously to eIF1 and eIF2␤-NTD. Together, these results support the idea that a trimeric eIF1⅐eIF2⅐eIF5 complex can be formed as a part of MFC for stimulation of TC binding to the ribosome. FLAG Tagging of eIF1 Impairs the Interaction with eIF5-CTD, eIF2␤-NTD, and the Native eIF3 Complex in Vitro-To analyze the functional significance of a protein interaction in vivo, it is important to obtain a mutation that specifically reduces it. During the course of our study, we noted that some of the interactions involving eIF1 were compromised by its epitope tagging at either terminus with the bulky highly charged FLAG peptide (DYKDDDDK). As shown in Fig. 2A, middle and bottom panels, C-terminally and N-terminally FLAG-tagged eIF1, designated eIF1-FL and FL-eIF1, respectively, had reduced interactions with GST-eIF2␤-N and GST-eIF5-B6 (lanes 3 and 6) but not with GST-eIF3c-N (lane 4). In addition, we found that both forms of FLAG-eIF1 had reduced interactions with the C-terminal HEAT domain of eIF4G2 (lane 5) (24) as well as with the C-terminal domain of eIF3a (GST-TIF32⌬1 (8); data not shown).
To examine the interaction of FLAG-eIF1 derivatives with native eIF3, we affinity-purified native HA epitope-tagged eIF3, attached it to anti-HA affinity resin, and allowed it to  (7). 35 S-eIF1 was expressed from pT7-SUI1 (Table I) with the TNT/T7 system (Promega). After washing the beads four times with binding buffer without milk, bound proteins were eluted in Laemmli buffer and separated by SDS-PAGE. Then 35 S-eIF1 was detected by autoradiography. The percentage of 35 S-eIF1 bound to the beads was quantitated and indicated below. Lane 1, 20 or 50% of input amount of 35 S-eIF1. C, GST fusion proteins (ϳ5 g) indicated across the top are attached to glutathione-Sepharose beads and bound to eIF1 (12 g) present in induced extracts of BL21(DE3) carrying pT7-SUI1 (I) or extracts mock-induced from BL21(DE3) carrying an empty vector pT7-7 (U). Binding and washing were done exactly as in A and B, and GST fusion protein complexes were eluted by glutathione. 75% of eluted fraction was analyzed by Coomassie staining (top panel). 5% of eluted GST-eIF3c complexes (lanes 5-7) and 25% of other complexes (lanes 3, 4, and 8 -13) were analyzed by immunoblotting with anti-eIF1 antibodies (bottom panel). Lanes 1 and 2, 1% of input amount of BL21(DE3) cell extracts used in the binding reaction. Lanes 7, 10, and 13, GST fusion complexes were treated with 10 g of RNase A prior to elution. D, FLAG-tagged eIF2 was purified from strain KAY42 and attached to anti-FLAG-affinity resin, as described (6). The resins attached to FLAG-eIF2 (ϩ) or none (Ϫ) were incubated with bacterial extracts induced or uninduced for r-eIF1 (8.4 g) as defined in C and washed extensively. The entire precipitated fractions were analyzed by SDS-PAGE and Western blotting using antibodies indicated to the right (lanes 5, 6, 8, and 9). The amount of factors was quantitated in reference to their amounts in WCE (40 g) from a wild type strain KAY146 (lane 3) and indicated at the bottom of each panel. Lanes 1 and 2, Coomassie staining pattern of proteins eluted from anti-FLAG affinity resin attached to none and FLAG-eIF2, respectively. Lanes 4 and 5, 2% input (In) amount of bacterial extracts used for binding assays. *, a bacterial protein that cross-reacts with anti-eIF3g antibodies. E, bridging experiment.  Fig. 2B, lane 2, the purified HA-eIF3 contains stoichiometric amounts of a, b, and g subunits of eIF3, a substoichiometric amount of eIF5, and no detectable eIF1. This eIF3 complex bound untagged eIF1 very efficiently but bound FLAG-eIF1s less efficiently. Thus, FLAG tagging of eIF1 reduces the interaction with eIF3.
FLAG Tagging of eIF1 Disrupts the Interaction with the eIF3⅐eIF2⅐eIF5 Complex in Vivo-To test the in vivo effect of FLAG tagging of eIF1, we constructed single copy plasmids YCpL-FL-SUI1 (FL-SUI1 LEU2) or YCpL-SUI1-FL (SUI1-FL LEU2) encoding FL-eIF1 or eIF1-FL, respectively (see Tables I  and III). The FL-eIF1 construct contains a Met residue inserted between the FLAG peptide and the second codon of eIF1. To avoid possible leaky scanning that might allow expression of an untagged eIF1 from the FL-eIF1 plasmid, we also constructed YCpL-FL-SUI1*, altering the second Met residue of FL-eIF1 to Leu. This eIF1 derivative was designated FL-eIF1*. To verify the expression of FLAG-eIF1 derivatives in vivo, we introduced the above mentioned plasmids to a wild-type yeast strain, Y217 (Table II) and analyzed the resulting transformants for immunoblotting. As shown in Fig. 3A, top panel, immunoblotting with anti-FLAG antibodies showed that YCpL-FL-SUI1 (lanes 1 and 2) and YCpL-FL-SUI1* (not shown) produced FLAG-eIF1 species migrating at 16 and 18 kDa, whereas YCpL-SUI1-FL produced eIF1-FL migrating at 14 kDa (lanes 8 and 9). We  (Table I)

eIF1 Recruitment to 40 S Ribosome via MFC
confirmed that these species indeed reacted with anti-eIF1 antibodies (Fig. 3A, bottom panel). The 16-kDa species of FL-eIF1 and the 14-kDa species of eIF1-FL comigrated with recombinant forms of FL-eIF1 and eIF1-FL (with predicted masses of 13.4 and 13.3 kDa, see Table III), respectively, indicating their anomaly in electrophoretic mobility. We do not know how the larger, 18-kDa FL-eIF1* or FL-eIF1 was produced. Finally, comparison with the amount of endogenous eIF1 (12 kDa) expressed from the unmodified SUI1 allele in Y217 allowed us to judge the expression level of the three FLAG-eIF1 derivatives to be equivalent to that of endogenous eIF1 (see Table III, column 4, for the results of quantitation).
To first test the abilities of the FLAG-tagged eIF1 to bind eIF3 in vivo, we introduced each of these plasmids or an empty vector to strain KAY107 encoding an HA-tagged form of eIF3i or isogenic KAY6 encoding an untagged eIF3i (Table II) and conducted immunoprecipitation with anti-HA antibodies. Note that these transformants express endogenous (untagged) eIF1 from the chromosome in addition to FLAG-tagged eIF1 encoded by the introduced plasmids. As shown in Fig. 2C, lanes 2 and 5, experiments with vector controls confirmed that endogenous eIF1 co-precipitated specifically with HA-eIF3i, together with eIF3b subunit, eIF2␣ subunit, and eIF5 as components of the MFC (4). In agreement with in vitro binding assays between native eIF3 and FLAG-eIF1 derivatives (Fig. 2B), little of the FLAG-eIF1 constructs immunoprecipitated with HA-eIF3 (Fig.  2C, bottom panel, lanes 8, 11, and 17). Thus, all three forms of FLAG-eIF1 cannot bind eIF3 efficiently in vivo.
Interestingly, we found that endogenous eIF1 was not associated with HA-eIF3 in the presence of eIF1-FL (Fig. 2C, fifth  panel, lane 17), although it was associated with HA-eIF3 in the presence of FL-eIF1* (lane 8) to the same degree as found in the vector control transformant (lane 5). Thus, eIF1-FL has a dominant negative effect on native eIF1 binding to eIF3, whereas FL-eIF1* does not. Because the in vitro binding assay in Fig.  2B indicates that eIF1-FL can bind HA-eIF3, albeit less efficiently than wild type eIF1, we presume that the weak association of eIF1-FL with eIF3-HA interfered with the latter's binding to endogenous eIF1. Perhaps, the washing step in the coimmunoprecipitation experiment was harsh enough to disrupt this weak interaction. These results are consistent with the more severe dominant negative effect of eIF1-FL overexpression (see Figs. 4

and 5).
FLAG Tagging of eIF1 Reduces the Interaction with the 40 S Ribosome in Vivo-We then tested whether the FLAG-tagged forms of eIF1 can be recruited to the 40 S ribosome in vivo. For this purpose, yeast strains carrying YCpL-FL-SUI1* or YCpL-SUI1-FL expressing FL-eIF1* or eIF1-FL, respectively, or an empty vector were grown exponentially and harvested after translation elongation was "frozen" by adding cycloheximide. Immunoblot analyses of sucrose gradient fractions prepared from the control vector transformant indicated that ϳ20% of endogenous eIF1 bound to the 40 S ribosome, together with similar proportions of eIF2␣, -3a, and -5 ( Fig. 4A; also see Table  IV below). Analogous experiments with the plasmid encoding FL-eIF1* (Fig. 4B) or eIF1-FL (Fig. 4C) indicated that smaller proportions of these species were found in the 40 S ribosomal fractions than that of endogenous eIF1 bound to the 40 S ribosome in the same cell extracts (Fig. 4, D and E, compare lane 2 versus lane 4). These results indicate that the FLAG tagging of eIF1 impairs its recruitment to the 40 S ribosome.
As summarized in Table IV, we consistently find 20 -30% of total MFC components (eIF1, -2, -3, and -5) in the 40 S ribosome fractions isolated from wild type yeast, and the molar levels of these proteins are roughly equivalent 3 (also see Ref. 18). These data are consistent with the idea that the 43 S complex contains stoichiometric amounts of MFC components, as a result of binding of MFC as a preformed unit (4), although they do not rule out the possibility that the MFC assembled on, rather than free of, the 40 S subunit.
C-terminal FLAG Tagging of eIF1 Is Lethal-Next, we wished to test the effect of FLAG tagging on in vivo function of eIF1. For this purpose, plasmids encoding differently FLAG-tagged forms of eIF1 were introduced to the SUI1 deletion strain Y217 harboring a URA3 SUI1 plasmid (Table  II), and the latter plasmid was removed by growth on medium supplemented with 5-fluoroorotic acid (FOA) that is toxic to cells expressing the Ura3p enzyme. As shown in Fig. 3B, line c, we found that the cells expressing eIF1-FL did not grow on the FOA medium, indicating that C-terminal FLAG tagging of eIF1 is lethal. Because very little eIF1-FL was associated with the 40 S ribosome (Fig. 4C), we suggest that the inability  Table I) were introduced into Y217 (sui1⌬ p͓URA3 SUI1͔). Expression levels were tested after evicting the URA3 SUI1 plasmid on FOA medium except for lethal alleles. Thus, the values under columns 4 and 5 are the relative amounts of eIF1 expressed solely from the plasmids under column 3. For lethal SUI1-FL alleles, expression level was examined in the presence of p͓URA3 SUI1͔ in Y217; since the level of untagged 12-kDa eIF1 species from these alleles was indistinguishable from the level of native eIF1 from the control SUI1 ϩ strain KAY146, the level of untagged eIF1 (with FLAG epitope removed proteolytically) in column 5 was judged to be zero. Growth was tested before (dominant, column 6) or after (recessive, column 7) evicting p[URA3 SUI1͔ in Y217.
a FL-eIF1 contains MDYKDDDDKM, followed by the natural eIF1 peptide from the second amino acid. FL-eIF1* replaces the second Met of FL-eIF1 with Leu. eIF1-FL carries DYKDDDDK, following the entire eIF1 peptide. Predicted molecular masses are 13,429, 13,411, and 13,298 Da, respectively, compared with 12,304 Da for wild type. Transcription of all these alleles starts from the natural SUI1 promoter, except for GPD FL-SUI1 starting from a stronger GPD promoter. b FLAG-tagged and untagged forms of eIF1 were detected by immnoblotting with anti-eIF1 antibodies (example shown in Fig. 3C). The relative amounts of these species per fixed amount of WCE were determined by densitometry with the NIHImage software and expressed in comparison with the amount of native eIF1 in wild type yeast (the value of 1 in parenthesis). c WT, normal growth; Lethal, no growth on FOA; Slg Ϫ , slow growth at 30 and 37°C. d Tested by introducing the plasmids into strain KAY37 (gcn2D) and checking for 3-AT resistance (Fig. 5). ϩ, no growth at 10 mM 3-AT; Ϫ, growth at 20 mM 3-AT; ϩ*, no growth at 10 mM 3-AT but derepressed GCN4 translation. eIF1 Recruitment to 40 S Ribosome via MFC of at least this protein to bind the ribosome produced the rate-limiting defect.
To test whether this lethal effect of eIF1-FL can be overcome by mass action, we constructed YEpL-SUI1-FL encoding eIF1-FL on a high copy vector. As shown in Fig. 2C, lanes 19 -21, co-immnoprecipitation of HA-eIF3 in KAY107 (HA-TIF34) transformant carrying this plasmid confirms that eIF1-FL associates with HA-eIF3 when overproduced. However, the SUI1-FL allele on this plasmid did not complement the SUI1 deletion (Fig. 3B, row g), although it expressed eIF1-FL at a 6 -8 times higher level than YCp-SUI1-FL (Table  III). These results indicate that increased gene dosage cannot complement the rate-limiting defect of eIF1-FL. Consistent with this idea, we found that the 43 S complex formation in the presence of excess eIF1-FL was severely inhibited, with fewer MFC components (eIF1, eIF2, eIF3, and eIF5) associated with the 40 S ribosome (Fig. 4F, lanes 7 and 8). Thus, excess eIF1-FL impedes 43 S complex formation rather than restoring it.
Together with the data in Fig. 2C, lanes 19 -21, these results suggest that excess eIF1-FL recovers complex formation with eIF3 by mass action, but this complex cannot be recruited to the 40 S ribosome. Therefore, C-terminal FLAG tagging might impair efficient binding of MFC to the 40 S ribosome in addition to impairing incorporation into the MFC (see "Discussion"). The severe inhibition of 43 S complex formation caused by eIF1-FL overexpression obviously impaired general translation initiation, because the Y217 transformant carrying YEpL-SUI1-FL grew slowly at all of the temperatures tested (hence a dominant slow growth phenotype), concomitant with a reduced polysome content, as shown in Fig. 3D.
In contrast to the effect of eIF1-FL, the cells expressing FL-eIF1 or FL-eIF1* grew on the FOA medium (Fig. 3B, lines d and e), indicating that these alleles are not lethal. Immunoblot analyses showed that the resultant FOA r strains KAY156 (FL-SUI1) and KAY178 (FL-SUI1*) expressed untagged eIF1 (12 kDa) besides FL-eIF1 (16 and 18 kDa) (Fig. 3C, lanes 1-9), although the strains were Ura Ϫ and therefore lacked the URA3 plasmid carrying the native eIF1 allele. Notably, untagged eIF1 was diminished but not completely eliminated in KAY178 encoding FL-eIF1* compared with that in KAY156 encoding FL-eIF1 (Fig. 3C, lanes 3-9). Because FL-eIF1* carries an amino acid substitution at the second Met between FLAG and eIF1 peptides, the generation of untagged eIF1 in KAY178 (FL-SUI1*) is probably due to proteolytic cleavage of the tag. This situation makes it difficult to judge any recessive phenotype (including lethality) of N-terminal FLAG tagging of eIF1.
Overexpression of eIF1-FL Confers a Dominant Gcd Ϫ Phenotype-Having observed severe inhibition of 43 S complex formation in a strain overproducing eIF1-FL (Fig. 4F), we set out to test the strain's Gcd Ϫ phenotype, the in vivo barometer for Met-tRNA i Met binding to the ribosome (25). The Gcd Ϫ phenotype occurs when the general amino acid control pathway is constitutively activated by the action of Gcn4 transcription activator for amino acid biosynthesis enzymes. GCN4 translation is controlled by its upstream open reading frames (uORFs). In nonstarved cells, GCN4 translation is repressed, because 40 eIF1 Recruitment to 40 S Ribosome via MFC S ribosomes that are bound to the GCN4 leader following uORF1 translation dissociate after they translate uORF4 efficiently. When cells are starved for amino acids, uncharged tRNA accumulates and activates eIF2 kinase Gcn2p. Phosphorylated eIF2 competitively inhibits the action of the guanine nucleotide exchange factor eIF2B, thereby blocking Met-tRNA i -Met delivery to the 40 S ribosomes. Under these conditions, 40 S ribosomes migrating down the GCN4 leader rebind TC only after they pass uORF4, resulting in efficient GCN4 translation. Gcd Ϫ mutations derepress GCN4 translation independent of eIF2 phosphorylation (hence action of Gcn2p), due to reduced Met-tRNA i Met binding to the ribosome. Therefore, excess eIF1-FL would be expected to induce GCN4 translation, since it severely impedes 43 S complex formation, as shown in Fig. 4F.
To test this possibility, we introduced a high copy eIF1-FL plasmid and control plasmids to a gcn2⌬ strain. As shown with the control vector transformant (Fig. 5A, line 1), the gcn2⌬ strain is sensitive to 3-AT, since it inhibits a histidine biosynthesis enzyme, and this inhibition cannot be rescued by activating the general control pathway due to gcn2⌬. As shown in Fig. 5A, line 7, the gcn2⌬ transformant carrying high copy eIF1-FL plasmid is now 3-AT-resistant, just as observed with the control transformant carrying high copy eIF5 plasmid (line 8). Overexpression of eIF5 sequesters eIF2 TC in a partially formed MFC by its direct interaction with the latter, thereby impeding 43 S complex formation (6). Note, however, that the 3-AT resistance observed with eIF5 and eIF1-FL overexpression is not as strong as the normal 3-AT resistance conferred by Gcn2p kinase (Fig. 5B). The Gcd Ϫ phenotype of the eIF1-FL transformant is due to C-terminal FLAG tagging of eIF1 but not due to eIF1 overexpression per se, because overexpression of native eIF1 did not suppress the 3-AT sensitivity of the gcn2⌬ strain (Fig. 5A, line 3). Importantly, the Gcd Ϫ phenotype of eIF1-FL transformant is due to limiting eIF2 TC binding to the ribosome, because co-overexpression of all three eIF2 subunits and tRNA i Met suppressed 3-AT resistance conferred by eIF1-FL overproduction as well as that caused by eIF5 overproduction (Fig. 5C, compare lanes 14 and 15 with lanes 17 and  18). Immunoblot analyses in Fig. 5D indicate that the reduction in 3-AT resistance by eIF2 TC overexpression occurred without altering the level of eIF1-FL. These results strongly support the physiological relevance of the inhibition of 43 S complex formation by excess eIF1-FL, as observed in Fig. 4F.
In contrast to the effect of eIF1-FL, we did not observe a dominant Gcd Ϫ phenotype with any FL-eIF1 or FL-eIF1* construct either on a single copy or high copy vector (Fig. 5A, lines 4 and 5; data not shown for FL-eIF1). Immunoblot analyses indicate that FL-eIF1* level from the high copy vector is ϳ3fold higher than eIF1-FL level from the same vector and that the level of untagged eIF1 expression from the high copy FL-eIF1* plasmid is low (Table III). Thus, we believe that the effect of FL-eIF1* on GCN4 expression is smaller than that of eIF1-FL. Alternatively, the residual expression of untagged eIF1 from the high copy FL-eIF1* plasmid may be sufficient to competitively diminish its effect on GCN4 expression. eIF1 Recruitment to 40 S Ribosome via MFC expresses FL-eIF1 4 -5-fold more than the level of eIF1 in wild type strains, as shown in Fig. 3C, lanes 10 -13. Due to the second Met codon between FLAG and eIF1 peptides, native eIF1 was also highly expressed in this strain (Fig. 3B, lanes  10 -13). Strain KAY142 grows normally at all of the temperatures tested (Table III) and displayed a normal polysome profile (Fig. 3E), indicating that general translation initiation is not inhibited by the FL-eIF1 overexpression.
As shown in Fig. 6B, sucrose gradient analyses of cell extracts prepared from KAY142 indicate that the level of eIF2 on the 40 S ribosome was modestly reduced compared with wild type KAY146 (Fig. 6A), without altering the ribosomal association of eIF3 (as detected by anti eIF3a and eIF3b antibodies) and eIF5. Densitometric measurements from three independent experiments demonstrated a 2-fold reduction in eIF2 association with 40 S ribosome (Fig. 6C, bottom panel) and no alteration of eIF3 and eIF5 association with 40 S ribosome (top to third panels). Consistent with the results shown in Fig. 4B, N-terminal FLAG tagging of eIF1 reduces the level of FL-eIF1 binding to the 40 S ribosome, compared with the native form expressed from the same allele, GPD FL-SUI1 (Fig. 6D). Together, these results indicate that FL-eIF1 overexpression partially impedes 43 S complex formation in vivo.
We believe that the 2-fold decrease in eIF2 binding to the ribosome is significant, because we found that GCN4 translation from a GCN4::lacZ reporter plasmid was partially derepressed in strain KAY142, consistent with this observation (data not shown). DISCUSSION eIF1 plays a critical role in accurate selection of AUG as a start codon as shown both by yeast genetics and mammalian biochemistry (see Introduction). In this report, our analyses of FLAG-tagged forms of eIF1 as mutants provided in vivo evidence for a second function of eIF1 in promoting the 43 S complex formation. Specifically, we found that FLAG-tagged forms of eIF1 reduce their binding to eIF3 in vitro (Fig. 2B), thereby reducing incorporation into the 43 S complex in vivo (Fig. 4, A-E). Importantly, C-terminal FLAG-eIF1 overexpression severely impedes the 43 S complex formation in vivo (Fig.  4F) and confers a dominant Gcd Ϫ phenotype that is suppressible by overexpression of eIF2 TC (Fig. 5). N-terminal FLAG-eIF1 overexpression also reduces binding of eIF2 to the 40 S ribosome (Fig. 6) and moderately derepresses GCN4 translation (data not shown). These results indicate that the incorporation of eIF1 into the MFC is critical for formation of 43 S complex in vivo. eIF1 was recently shown to stimulate eIF2 TC binding to the 40 S ribosome in vitro both in yeast (13) and mammals (14). Together with our in vivo data, these data strongly support the second role of eIF1 in promoting 43 S complex assembly. The dual role for a single factor in 43 S complex formation and accurate AUG selection was first proposed for eIF5, the GAP for eIF2 (4,5).
How does eIF1 promote the 43 S complex formation? eIF1 appears to be a part of nucleation site for the MFC formation, composed of eIF2␤-NTD, eIF3c-NTD, eIF5-CTD, in addition to eIF1 itself (see Fig. 1G). Consistent with this idea, we provided biochemical evidence indicating that a trimeric complex can be formed between eIF1, eIF2␤-NTD, and eIF5-B6 (Fig. 1, E and  F). Because eIF3 and eIF2 can bind directly (8) and have independent ribosome-binding sites (26), eIF1 should be able to facilitate 43 S complex formation in a cooperative manner, in concert with the action of eIF5-CTD. We believe that the ribosome-binding defect observed with C-terminal and N-terminal FLAG-eIF1 (Fig. 4, B, C, E, and F) is due to severing eIF1 from eIF2, eIF5, and eIF3 (Fig. 2), thereby impairing this mechanism of 43 S complex promotion by eIF1.
A second, less established mechanism for eIF1 to promote 43 S complex formation would be to assume that eIF1 allosterically activates either MFC or the 40 S ribosome for 43 S complex formation. The idea that eIF1 might allosterically control eIF2 TC binding to the 40 S ribosome is not new but has yet to be proven (23). In this study, we observed that overexpression of C-terminal FLAG-eIF1 can restore binding of the mutant eIF1 to eIF3 (Fig. 2C, lanes 19 -21), but the MFC formed with the mutant eIF1 appears to be defective in binding to the 40 S ribosome (Fig. 4F). Because it was shown recently that eIF1 can directly bind to the 40 S ribosome in vitro at a high affinity (27), one might suspect that this direct contact is the major driving force for MFC binding to the ribosome and is inhibited by eIF1-FL. However, we believe that this is not the case, because the prt1-1 mutation altering the eIF3b subunit impairs eIF1 as well as eIF3 and TC binding to the 40 S ribosome in cell extracts, indicating that the recruitment of MFC constituents, including eIF1, depends on eIF3 (10). If so, it would be attractive to propose that eIF1 allosterically stimulates the ability of other factors (likely eIF3) to bind the 40 S ribosome and that C-terminal FLAG tagging impairs this function, in addition to impairing incorporation into the MFC.
Indeed, a similar role in MFC binding to the ribosome could be proposed for eIF5, since it enhances the ability of the eIF3a⅐eIF3c subcomplex to bind the 40 S ribosome (26).
The effect of FLAG tagging of eIF1 on the 43 S complex formation appears to be different between N-terminal and Cterminal FLAG-eIF1 (Figs. 2-6), although both forms lower the affinity with individual MFC partners in a similar manner ( Fig. 2A). This difference may arise from different locations of the tag introduced, since the C terminus of eIF1 ends in an antiparallel ␤-sheet, whereas its N-terminal ϳ20-amino acidlong segment is unfolded and contains conserved hydrophobic/ acidic residues, suggestive of interaction with other partners (28). The antiparallel ␤-sheet domain of eIF1 may be important for MFC binding to the ribosome, either through direct contact with the ribosome as proposed previously (27) or through somehow activating eIF3, as suggested above.
In the case of N-terminal FLAG-eIF1 (FL-eIF1 or FL-eIF1*), the in vitro binding defect with eIF3 appears to be more severe than the defect caused by eIF1-FL (Fig. 2B, compare lanes 6   FIG. 6. FL-eIF1 overexpression partially inhibits 43 S complex formation in vivo. Strains KAY142 ( GPD FL-SUI1) (B) and its isogenic wild type KAY146 (A) were grown in YPD and subjected to polysome analysis exactly as in Fig. 4. Lane 1, 5% of input amount of cell extracts. Lanes 2-11, 50% of sucrose gradient fractions. C, eIF distribution between free and 40 S fractions. Amounts of eIF3a, eIF3b, eIF5, and eIF2␣ in combined fractions 2-7 (free) or 8 and 9 (40 S) were compared with the total amount of each factor in lanes 2-11. We compared with this amount rather than the input amount in lane 1 to avoid larger S.D. values (see Table IV). D, amounts of untagged eIF1 and FL-eIF1 in free and 40 S fractions from KAY142 were quantitated exactly as in Fig. 4, D and E.

TABLE IV
Relative eIF levels in free and 40 S ribosome fractions in yeast Wild type strain KAY146 was grown to A 600 ‫ف‬ 1 at 30°C in YPD and harvested after cycloheximide was added to the medium for 5 min. WCEs prepared from this strain were subjected to sucrose gradient followed by immunoblotting as described under ''Materials and Methods.'' An example of immunoblot is shown in Fig. 6A. The amount of factors found in lanes 8 and 9 (40 S fraction, column 3) and lanes 2-7 (free fraction, column 4) was quantitated using NIHImage software and compared with the input amount in lane 1. Given in parentheses are S.D. values from three independent experiments. Column 2 indicates relative level of the corresponding factor according to our and 8). We consistently observed that FL-eIF1 was not incorporated into the MFC in vivo (Fig. 2C). Thus, the N-terminal portion of eIF1 may be more critical for incorporation into the MFC. Despite the more severe defect in eIF3 binding, FL-eIF1 overexpression from the GPD promoter, resulted in a minor decrease in the level of eIF2 on the 40 S ribosome (Fig. 6) and derepressed GCN4 translation (data not shown; also see Ref 21). A possible explanation for these findings might be to propose that FL-eIF1 separately interacted with eIF2 by mass action, thereby sequestering a part of eIF2 from binding to the ribosome. It is possible that the inhibition of TC binding in this manner was very moderate and resulted in only a minor increase in GCN4 expression. Consistent with a defect in eIF3 binding (Fig. 2C) and possibly an additional defect in MFC binding to the ribosome (Fig.  4F), the SUI1-FL allele encoding eIF1-FL was recessive lethal, regardless of the copy number of the vector (Fig. 3B and Table  III). We could not judge the recessive phenotypes of N-terminal FLAG-eIF1-encoding alleles, because untagged eIF1 was produced from the same expression plasmids, possibly due to proteolytic cleavage of the tag (Fig. 3C). Therefore, it would be important to generate and characterize eIF1 mutant derivatives that cannot bind MFC (like FL-eIF1 or FL-eIF1*) but do not produce proteolytic derivatives that might complement the phenotypes produced by the mutations themselves.