Functional Characterization of the Role of the N-terminal Domain of the c/Nip1 Subunit of Eukaryotic Initiation Factor 3 (eIF3) in AUG Recognition^*

Construction of Yeast Strains and Plasmids

HMJ04 and HMJ06 were generated by introducing YCpMJ-MET-Nip1-W into HLV04 and HKN06 (11), respectively. The original pNIP1⁺ (c/NIP1 URA3) covering plasmid was evicted on SD plates containing 5-fluoro-orotic acid.

To create HMJ08, H2880 (21) was first transformed with YCpNIP1-His-L to cover for the deletion of NIP1 that was made in the next step by introducing the SacI-SphI fragment carrying the nip1Δ::hisG-URA3-hisG integration cassette from pLV10 (11). The uracil auxotrophy was regained by growing the cells on SD plates containing 5-fluoro-orotic acid. The resulting strain was subsequently transformed with YCpNIP1-His-U, and the leucine auxotrophy was regained by growing the cells in liquid medium containing leucine and selecting for those that lost the YCpNIP1-His-L plasmid on SD ± leucine plates producing HMJ08.

YCpMJ-MET-NIP1-W was constructed by inserting the 2618-bp BamHI-HindIII fragment from pGAD-NIP1 (22) into YCplac22MET-W (23) digested by BamHI-HindIII.

YCpNIP1-Box12-WW was made by fusion PCR using the template YCpNIP1-His-L and two sets of primers (JKNIP1Mut and MJreversV111WV112W; LV22ext and MJ111WW) in the first PCR. The PCR products thus obtained were mixed in 1:1 ratio, and the second PCR was performed using primers JKNIP1Mut and LV22ext. The resulting PCR product (631 bp) was cut with AvaI-XbaI and inserted into AvaI-XbaI-digested YCpNIP1-His-help3 (11).

The following plasmids were created essentially the same as those described above with the exception of the primer sets (containing the degenerate oligonucleotides) that were used in the first PCRs: YCpNIP1-Box12-SPW (JKNIP1Mut and LV12C-1XR; LV22ext and LVBOX12C-1X), YCpNIP1-Box12-WWPW (JKNIP1Mut and LV12C-3XR; LV22ext and LVBOX12C-3X), YCpNIP1-GAP⁸⁵ (JKNIP1Mut and LV12C-2XR; LV22ext and LVBOX12C-2X), YCpNIP1-GAP⁹² (JKNIP1Mut and LV12C-4XR; LV22ext and LVBOX12C-4X), and YCpNIP1-Box15-2 (JKNIP1Mut and LV15C-2XR; LV22ext and LVBOX15C-2X).

YEpNIP1-GAP⁸⁵, -GAP⁹², and -Box15-2 were constructed by inserting the 2887-bp HindIII-PstI fragments from the corresponding single copy plasmids described above into YEplac181 digested with HindIII-PstI. pT7-NIP1-N270, pT7-NIP1-N-Box12-SPW, pT7-NIP1-N-Box12-WWPW, pT7-NIP1-N-Box12-WW, pT7-NIP1-N-Box15-2, pT7-NIP1-N-GAP⁸⁵, and pT7-NIP1-N-GAP⁹² were all made by inserting the NdeI-HindIII-digested PCR product obtained with primers JKNIP1Mut and LVRN270 using the templates YCpNIP1-His-L, YCpNIP1-Box12-SPW, YCpNIP1-Box12-WWPW, YCpNIP1-Box12-WW, YEpNIP1-Box15-2, YEpNIP1-GAP⁸⁵, and YEpNIP1-GAP⁹², respectively, into NdeI-HindIII-digested pT7-7.

pT7-NIP1-N-Δ46/137 and pT7-NIP1-N-Δ60 were created by inserting the NdeI-BamHI-digested PCR product obtained using the template YCpNIP1-His-L and the set of primers MJNIP1_46–137 and MJNIP1_46–137rev or MJNIP1_61–205 and MJNIP1_61–205rev, respectively, into NdeI-BamHI-digested pT7-7. Lists of all yeast strains and plasmids can be found in Tables 1 and 2, respectively; a list of all PCR primers will be provided upon request.

Yeast Biochemical Methods

GST pulldown experiments with GST fusions and in vitro synthesized ³⁵S-labeled polypeptides (see Table 2 for vector descriptions) were conducted as described in Fig. 5. Ni²⁺ chelation chromatography of eIF3 complexes containing His₈-tagged c/Nip1 from yeast whole-cell extracts (WCEs) and Western blot analysis were conducted as described in detail previously (24). In short, WCEs were incubated at 4 °C overnight with 15 μl of 50% Ni²⁺-Sepharose 6 Fast Flow (GE Healthcare) suspended in 300 μl of buffer A (20 mm Tris-HCl (pH 7.5), 100 mm KCl, 5 mm MgCl₂, 0.5 mm β-mercaptoethanol, 1 mm phenylmethylsulfonyl fluoride, 20 mm imidazole, 10% (v/v) glycerol, 1 EDTA-free complete Protease Inhibitor Mix tablets (Roche)) followed by washing and elution. Polysome profile analysis and 2% HCHO cross-linking followed by WCE preparation and fractionation of extracts for analysis of pre-initiation complexes were carried out as described previously (24). β-Galactosidase assays were conducted as described (25).

Semirandom Mutagenesis of the Selected Segments of the N-terminal Domain of the c/Nip1 Subunit of eIF3

The c/Nip1-NTD mediates eIF3 interactions with eIFs 1 and 5 and indirectly with eIF2 (20). By subjecting the N-terminal 160 amino acids of c/Nip1 to the clustered 10-alanine mutagenesis, we showed previously that these c/Nip1 interactions stimulate the assembly of the 43 S PICs and somehow coordinate the functions of eIF1 and eIF5 in stringent AUG selection (11). To understand the molecular mechanism of the c/Nip1 involvement in these functions and also to identify specific residues that are critical for them, we randomly mutated selected groups of conserved residues within the previously identified clustered 10-alanine mutagenesis c/nip1-Box mutations 6R, 12, and 15 (Fig. 2A) (20) using degenerate oligonucleotides and the fusion PCR technique. Degenerate oligonucleotides were designed to substitute a given group of amino acid residues with degenerate codons NNN where the N nucleotide was synthesized with a mixture of A, C, G, and T, each at 25% of the total. This way we constructed nine mutant libraries containing random amino acid residues at selected sites within each group derived from the latter three Boxes (Fig. 2A). These mutant libraries were then separately introduced into two different yeast strains (both deleted for chromosomal NIP1 and carrying a wild-type (WT) NIP1 allele under the control of MET3 promoter) and screened for the Sui⁻ and Gcd⁻ phenotypes indicative of a relaxed stringency of AUG recognition (strain HMJ04) or of a defect in 43 S PIC assembly (strain HMJ06), respectively, in the presence of methionine, which shuts off expression of the WT gene.

Mutations that relax the stringency of translational start site selection in yeast (so-called Sui⁻ mutations) were isolated by selecting for growth on medium lacking histidine in a his4-303 genetic background where the start codon of the HIS4 gene has been mutated and instead its third coding triplet (UUG) is used to initiate HIS4 translation (26). In other words, such mutations suppress histidine auxotrophy of the WT his4-303 cells, turning their His⁻ into a His⁺ phenotype. Mutations that affect formation of 43 S PICs by reducing the rate of the TC recruitment (so called Gcd⁻ mutations) were isolated by their ability to constitutively derepress GCN4 translation in cells lacking the kinase Gcn2, which makes them resistant to the inhibitor of histidine biosynthesis, 3-aminotriazole (3-AT). Gcn4 is a transcriptional activator of amino acid biosynthetic genes whose expression is under tight translational control via four short upstream ORFs (uORFs) occurring in its leader mRNA. The gcn2Δ cells in which translational activation of GCN4 is eliminated fail to grow on 3-AT-containing medium (for more details, see Ref. 27).

Approximately 10,000 clones from each library were screened individually for the His⁺ and 3-AT^Res phenotypes. Altogether, five mutants with strong phenotypes were isolated, two of which fall in the Box12 group and one that falls in the Box15 group (Fig. 2B). In addition to those, a few other mutations in the latter Boxes were also identified; however, given their less pronounced phenotypes, we did not characterize them any further. Similarly, no mutations with strong phenotypes were found in the Box6R group; hence, we stopped working with this group at this point. Although not programmed intentionally, two large internal deletions of 85 (GAP⁸⁵) and 92 (GAP⁹²) residues impinging into the region between Asp⁴⁶ and Asn¹⁴⁴ of the c/Nip1-NTD also resulted from this mutagenic procedure (Fig. 2B). All mutant phenotypes were corroborated to be plasmid-born by retransformation of isolated plasmid DNAs into the nip1Δ strains carrying the WT NIP1 URA3 covering plasmid that was subsequently evicted by plasmid shuffling. As shown in Fig. 2, C and D (summarized in Fig. 2B), three major classes of mutant substitutions were identified, imparting either (a) only the Sui⁻ phenotype (c/nip1-Box12-SPW (Fig. 2C, lane 3)), (b) only the Gcd⁻ phenotype (c/nip1-Box15-2 (Fig. 2D, lane 7)), or (c) both combined (c/nip1-Box12-WWPW and both GAP mutants (Fig. 2, C and D, lanes 4–6, respectively)). All of these mutations also resulted in a slow growth (Slg⁻) phenotype of different strengths as illustrated by their varying effects on translation initiation rates estimated by measuring the polysome:monosome ratios (summarized in Fig. 2B).

Our finding that no Sui⁻ mutations were detected in the Box15 group is consistent with the fact that the original c/nip1-Box15 also displayed only the Gcd⁻ phenotype (11). In contrast, whereas the original c/nip1-Box12 mutation produced merely the mild Sui⁻ phenotype, our semirandom substitutions of Box12 impart strong Sui⁻ as well as Gcd⁻ phenotypes. Perhaps the most striking observation, however, is the fact that whereas the deletion of the first 160 residues is lethal (11) the internal deletions of 84 or 91 residues are not lethal and resulted “only” in severe Slg⁻, Sui⁻, and Gcd⁻ phenotypes. Because the protein levels of Box15-2 and both GAP mutants in a single copy (sc) number were found partially reduced when compared with those of the WT c/Nip1, we generated high copy number versions of these three alleles. High copy (hc) expression increased their protein levels by ∼2-fold and more or less matched the sc expression of the WT c/Nip1 (Fig. 2E). Importantly, because there was no qualitative difference in the mutant phenotypes between the sc and hc expressions of the latter alleles (data not shown), we decided to use their hc versions throughout the rest of the study.

Dissection of Effects of the Specific c/Nip1-Box12 Residues on Phenotypes Indicating Impaired Formation of PICs Versus AUG Selection

We noted that substitution of three Box12 lysines (113, 116, and 118) by Ser, Pro, and Trp in c/nip1-Box12-SPW produced merely the Sui⁻ phenotype, whereas substitution of two preceding highly conserved valines (111 and 112) by tryptophans in combination with K116P and K118W in c/nip1-Box12-WWPW produced both the Sui⁻ and Gcd⁻ phenotypes (Fig. 2, B–D). The fact that both mutants contain K116P and K118W substitutions might thus suggest that these two also highly conserved lysines specifically contribute to the stringency of AUG selection, whereas Val¹¹¹ and Val¹¹² participate in assembly of the PICs. To test that, we separated the V111W and V112W double substitution from K116P and K118W by producing two individual c/nip1-Box12-WW and c/nip1-Box12-PW mutant alleles, which were then tested for Sui⁻ and Gcd⁻ phenotypes as described above. Strikingly, in accord with our prediction, c/nip1-Box12-WW imparts the severe Gcd⁻ phenotype but not Sui⁻ (Fig. 2D, lane 8). At odds with our prediction, the c/nip1-Box12-PW mutant showed no apparent phenotype (data not shown). Thus, we propose that whereas the two consecutive valines (Val¹¹¹ and Val¹¹²) somehow promote recruitment of the TC to the small ribosomal subunit all three consecutive lysines (Lys¹¹³, Lys¹¹⁶, and Lys¹¹⁸) are critically required for proper detection of the AUG start codon. In support, one of the “weaker” Box12 mutants that was isolated in our initial screen but not analyzed further carried substitutions of exactly these three lysines (K113Y, K116H, and K118H) and displayed the Sui⁻ but not Gcd⁻ phenotype (data not shown).

To demonstrate that the Sui⁻ phenotype is indeed caused by increased selection of the UUG triplet as the start site and that our Sui⁻ mutants increase initiation rates at near-cognate codons in general, we examined expression of matched HIS4-lacZ reporters containing UUG or AUU versus AUG start codons (Fig. 3A). Previously, we reported that the original c/nip1-Box12 10-Ala substitution increased the UUG/AUG and AUU/AUG initiation ratios by ∼3- and 5-fold, respectively (11). In comparison, all three Sui⁻ mutants (12-SPW, -WWPW, and GAP⁸⁵) generated here conferred a much more dramatic increase in the UUG/AUG initiation ratio (from 12- to 29-fold) and a similar increase in the AUU/AUG initiation ratio (∼6-fold), indicating a strong preference for selecting UUG over the AUU as the false start site. As expected, the “Gcd⁻-only” Box12-WW mutant showed little to no increase in both ratios.

High Dosage Suppressor Analysis of the c/Nip1-NTD Sui⁻ Mutants by Selected eIFs Known to Participate in the AUG Recognition Process

We showed previously that the Sui⁻ phenotype of the original Box12 mutation was suppressed by overexpressing WT eIF1 despite the fact that at the same time its Slg⁻ phenotype was exacerbated (11). In contrast, overproduction of eIF5 had an intensifying effect on both Sui⁻ and Slg⁻ phenotypes of nip1-Box12. To examine genetic interactions between our new Sui⁻ mutants and these key players in the AUG recognition pathway, we transformed the nip1-Box12 mutants and nip1-GAP⁸⁵ with hc vectors overexpressing eIF1 or eIF5 and scored the resulting transformants for growth in the presence or absence of histidine (Fig. 3B).

In analogy with our original data, overexpressing eIF1 reduced the growth rate of both Box12 mutants in the presence of histidine and suppressed their Sui⁻ phenotype in its absence (Fig. 3B, panels ii and iii, lanes 5 and 2). Because the reduced growth rate of both mutants on His+ medium somewhat complicates our conclusion regarding the suppression of the Sui⁻ defect, we measured the UUG/AUG and AUU/AUG ratios in WWPW and SPW mutants bearing either empty vector or hc eIF1. We found that overexpression of eIF1 reduced the latter ratios in the SPW mutant by ∼44 and 40%, respectively, and in the WWPW mutant by ∼30 and 33%, respectively. These findings clarify that the eIF1-mediated suppression is significant but not full. Interestingly, high dosage eIF1 also suppressed the Sui⁻ phenotype of nip1-GAP⁸⁵; however, it did not worsen but improved its Slg⁻ phenotype (panel iv, lane 5), indicating that in this particular mutant the eIF1 suppression effect is more potent. Overexpression of eIF5 surprisingly had no significant effect on the GAP⁸⁵ mutant (panel iv, lane 6), whereas it exacerbated the Sui⁻ but not Slg⁻ phenotype of Box12-WWPW (panel ii, lane 6). It is worth reiterating that these two mutants are not only Sui⁻ but also display the Gcd⁻ phenotype, which may contribute to the observed genetic interactions that differ from the original “Sui⁻-only” Box12 mutant. Consistently, overexpressing eIF5 in the Sui⁻-only nip1-Box12-SPW allele had an intensifying effect on both its Slg⁻ and Sui⁻ phenotypes (panel iii, lane 3). The intensifying effect on the Sui⁻ phenotype is inferred from the fact that it reduced growth on His+ but not on His− medium. It should be mentioned that neither of these Sui⁻ mutants conferred a significant Ssu⁻ phenotype when combined with dominant negative Sui⁻ alleles of eIF5 (in SUI5^G31R) or SUI3 (a subunit of eIF2; in SUI3^S264Y) (data not shown). Taken together, it is evident that each of these three Sui⁻ mutants impairs the AUG selection process in a more or less different way (see “Discussion”).

The nip1-Box15-2, -Box12-WW, and -Box12-WWPW Mutants Affect Recruitment of the TC to the PICs in a Manner Partially Reversible by Overexpressing TC

The NTD of c/Nip1 indirectly interacts with the TC via their common binding partner eIF5 and thus promotes its recruitment to the 40 S ribosomes (11, 19, 20). As hinted above, an inability to recruit and/or stably anchor the TC to the PICs constitutively derepresses the otherwise tightly controlled expression of GCN4 under non-starvation conditions, producing the Gcd⁻ phenotype even in the absence of the Gcn2 kinase. Mutants having such a mechanistic defect in the gcn2Δ background are commonly characterized by having their Gcd⁻ defect suppressible by overexpression of the TC as an increased cellular concentration of TC offsets the defect in its recruitment. The failure of hc TC to suppress the Gcd⁻ phenotype suggests a more complex defect(s) in the GCN4 regulation, the nature of which has not been understood so far.

To determine which of our newly generated NIP1 mutants affect the TC loading to the PICs as the rate-limiting defect, we introduced them with either empty vector or hc TC and tested the resulting transformants for growth in the presence of 3-AT. As shown in Fig. 4A, hc TC significantly suppressed the Gcd⁻ phenotypes of the nip1-Box12-WW and -Box15-2 mutants and to a smaller degree perhaps that of -Box12-WWPW but not that of the GAP⁸⁵ mutant, although it did partially suppress its slow growth defect. Importantly, increased TC dosage also suppressed, in fact almost fully, the Slg⁻ phenotype of Box15-2, strongly indicating that the TC recruitment is the dominating defect in this mutant.

To quantify the Gcd⁻ phenotypes of all Gcd⁻ mutants and to confirm or exclude its suppression by hc TC in WWPW and GAP⁸⁵, we assayed expression of a GCN4-lacZ reporter containing all four uORFs in the mRNA leader under non-starvation conditions (Fig. 4B). As expected, the WT gcn2Δ NIP1⁺ strain showed constitutively low GCN4-lacZ expression irrespective of the presence of either empty vector or hc TC, whereas all four mutations bearing an empty vector displayed significantly increased derepression of GCN4-lacZ expression in the gcn2Δ background after normalizing for their effects on the expression of an incontrollable GCN4-lacZ construct lacking all four uORFs (Fig. 4B). These results thus confirm that all of our Gcd⁻ mutants do diminish translational repression of GCN4 imposed by its uORFs as expected. Measuring the β-galactosidase activities in the presence of hc TC also confirmed moderate (by ∼30%) suppression of the WWPW Gcd⁻ defect and no suppression of that displayed by the GAP⁸⁵ mutant (Fig. 4B). Together, these results clearly imply that the impaired TC recruitment and thus the proper assembly of the PICs are to a varying degree one of the rate-limiting defects in all but one of our Gcd⁻ mutants.

The Extreme N Terminus of c/Nip1 Interacts with eIF5, the Segment That Immediately Follows Binds eIF1, and nip1-Box12-SPW Increases Binding Affinity of the c/Nip1-NTD for eIF1 in Vitro

Having analyzed all newly generated mutations by genetic means, we next sought to investigate molecular effects of the semirandom substitutions as well as internal deletions of the c/Nip1-NTD on binding affinities toward its direct binding partners in eIF1 and eIF5. We began by examining the effects of both GAP deletions on in vitro binding of ³⁵S-labeled c/Nip1-NTD (residues 1–270) to GST-eIF5 or GST-eIF1 fusions produced in Escherichia coli (Fig. 5A). Although this assay is more qualitative than quantitative in its nature, it has been reliably used in the past to provide rough estimates of altered binding affinities of mutant proteins (for example, see Refs. 9–13, 20, 24, and 28–30). In agreement with previous results, the WT c/Nip1-NTD polypeptide bound specifically to GST-eIF5 and GST-eIF1 but not to GST alone (Fig. 5B). Strikingly, both GAP deletions (Δ60–144 and Δ46–137) completely eliminated binding to eIF1 but not to eIF5 (lane 3 versus lane 4). In fact, binding to eIF5 was even strengthened perhaps because of the fact that the eIF5-binding site in the remaining c/Nip1 polypeptide was more ideally exposed for the interaction. In any case, given that the minimal c/Nip1 fragment required and sufficient for WT binding to both eIF1 and eIF5 is 156 residues (19), these results strongly suggest that the extreme N-terminal 45 residues constitute the binding site for eIF5, whereas the segment encompassing residues 60–137 represents the eIF1-binding site (Fig. 5A). In support, removal of the first 60 residues (in Δ60) or expressing the internal “GAP⁹²” segment (in Δ45/137) had the opposite effect; i.e. it eliminated binding to eIF5 but not to eIF1 (lane 4 versus lane 3). In an effort to directly demonstrate that the first 45 residues are sufficient for the interaction with eIF5, we created pT7 constructs expressing only the first 45 or 60 N-terminal residues of c/Nip1; however, they did not yield stable proteins. Taken together, these data suggest that the molecular nature of defects displayed by both GAPs is associated with the loss of eIF1 binding to the c/Nip1-NTD. This is consistent with the fact that hc eIF1 suppressed the Sui⁻ phenotype and partially suppressed Slg⁻ phenotype of GAP⁸⁵ (Fig. 3B, panel iv). Further analysis of binding affinities of semirandom Box mutants did not reveal any specific changes with the exception of Box12-SPW, which showed a reproducible ∼2-fold increase in binding to eIF1 (Fig. 5C, third panel, lane 3 with asterisk) (see also below). Note that Box12 falls in the eIF1 binding region, whereas Box15 lies further downstream.

nip1-GAP⁸⁵ Completely Diminishes Whereas nip1-Box12-SPW Modestly Increases the Amount of eIF1 Associated with the MFC in Vivo

To test whether our mutations affect binding of c/Nip1 to eIFs 1 and 5 and the TC in the context of the entire eIF3, we analyzed the in vivo composition of the MFC in cells expressing the individual His₈-tagged NIP1 mutant alleles by Ni²⁺ chelation chromatography. As reported previously (11), a fraction of a/Tif32, i/Tif34, g/Tif35, eIF2, eIF5, and eIF1 co-purified specifically with WT c/Nip1-His but not with its untagged version (Fig. 6A, lane 5 versus lane 2). Although neither of the mutations affects the extreme N-terminal binding site for eIF5, Box15-2 and GAP⁸⁵ reduced in vivo association of mutant eIF3 with eIF5 by more than 60% (lanes 17 and 20). In addition, Box12-WWPW also lowered the eIF3 binding affinity for eIF5 but to a smaller extent (by ∼30%) (lane 11). This is most probably an indirect effect arising from the overall changes of the c/Nip1-NTD fold in the context of the entire eIF3/MFC. As eIF5 mediates an indirect contact between eIF2 and the c/Nip1-NTD (19, 20), dramatic reductions in eIF2 association with the rest of MFC were also observed in these three mutant strains in a remarkable proportional accord with the extent of the eIF5 loss. In perfect agreement, all these NIP1 mutant alleles impart severe Gcd⁻ phenotypes (Fig. 2D), strongly suggesting that impaired TC loading to the PICs represents one of their functional defects in translation. Interestingly, another Gcd⁻ mutant, Box12-WW, showed virtually WT levels of eIF5 associated with the rest of eIF3 but significantly reduced amounts of eIF2 (lane 14). These results suggest that its TC binding defect mechanistically differs from those of the other three mutants (see “Discussion” for more details). In contrast, Box12-SPW, which is the only mutant allele that does not produce the Gcd⁻ phenotype, showed WT levels of both eIF5 and eIF2 (Fig. 6A, lane 8). This Sui⁻ mutant, however, did display a modest increase in the eIF1 amounts associated with the MFC, consistent with our in vitro binding data presented in Fig. 5C. It should be noted here that despite numerous repetitions the experimental error for eIF1 remained relatively high in this particular experiment. Nevertheless, because the increase in eIF1 recovery averaged out to ∼20% relative to WT, we propose that Box12-SPW slightly but specifically increases the binding affinity of the c/Nip1-NTD for eIF1 both in vitro and in vivo. Conversely, eIF1 binding was completely diminished by GAP⁸⁵ (Fig. 6A, lane 20) as expected because this mutant has the eIF1-binding site deleted. Finally, eIF1 binding in the remaining three Box mutants (12-WWPW, 12-WW, and 15-2) was also reduced but to a much smaller degree than in the case of the GAP⁸⁵ mutant. Taking our in vitro binding data into account (Fig. 5C), we think that these reductions are not a result of impaired direct binding between eIF1 and the NTD of c/Nip1 but rather an indirect consequence of changes in the arrangement of the assembly of the eIFs around the mutant c/Nip1-NTD.

The c/Nip1-NTD Mutants Generally Reduce 40 S Subunit Association of eIF3 and eIF5 in Vivo but Have Varying Effects on eIF1 and TC Recruitment to the 43–48 S PICs

All results described so far suggest that the three identified classes of the c/Nip1-NTD mutants that we generated have a varying impact on the composition of the 43–48 S PICs, the analysis of which would greatly help with the explanation of their molecular defects. To address this issue, we measured binding of selected eIF3 subunits and other MFC components to 40 S subunits by formaldehyde cross-linking followed by high velocity sedimentation in sucrose gradients (24). This method provides the best available approximation of the native composition of 43–48 S pre-initiation complexes in vivo. As shown in Fig. 7, A–C (left panels), we observed the expected co-sedimentation of a proportion of eIF3, eIF2, eIF5, and eIF1 with the 40 S species in the WCEs derived from WT cells. In the case of the class “a” Box12-SPW mutant, we observed a relative ∼50% decrease in the amounts of selected eIF3 subunits (c/Nip1, i/Tif34, and g/Tif35) and eIF5 associated with 40 S subunits (Fig. 7A, right panel), indicating that this mutation destabilizes eIF3 and its tightly bound partner eIF5 (28) on the ribosome in vivo perhaps by increasing their dissociation rates as suggested previously for some other eIF3 mutants (13, 31). In contrast, eIF2 levels remained practically unchanged, which nicely correlates with the fact that this mutant does not impart the Gcd⁻ phenotype. Importantly and in accord with our binding data mentioned above, we reproducibly detected significantly increased amounts of eIF1 in the 40 S subunit-containing fractions (note the peak of the eIF1 signal in the 40S subunit-containing fractions that is apparent only in the panel showing the SPW mutant). Hence, we propose that tighter binding of eIF1 to the c/Nip1-NTD somehow prevents eIF1 from properly functioning in the AUG start codon selection process, producing the severe Sui⁻ phenotype (see “Discussion” for our model).

The class “b” WWPW mutant imparting both the Sui⁻ and Gcd⁻ phenotypes reduced 40 S subunit binding of eIF3 to a similar extent and that of eIF5 to an even greater extent than did Box12-SPW (Fig. 7B, right panel). Here, the reduction in eIF5 on the 40 S ribosome might be attributed either to the reduced amount of eIF5 in the affinity-purified MFC and/or to the decrease in 40 S subunit-associated eIF3; we cannot rule out either of these possibilities. In accord with its Gcd⁻ phenotype and in contrast to Box12-SPW, it also significantly reduced the eIF2 levels on the ribosome. Another striking difference between WWPW and SPW mutants is that the former also markedly reduced the 40 S subunit-associated amounts of eIF1. These results thus imply that (i) the nature of the Sui⁻ defect differs between these two mutants and (ii) the character of the WWPW Sui⁻ defect conspicuously resembles so-called class “i” mutations in SUI1 encoding eIF1 (3). Class i Sui⁻ mutations weaken eIF1 association with PICs, and their Sui⁻ phenotype can be partially suppressed by their own overexpression; in accord, the Sui⁻ defect of WWPW is also partially suppressible by hc eIF1 (Fig. 3). Intuitively, the same results could be also expected with another class b mutant, GAP⁸⁵, which completely removes the eIF1-binding site (Fig. 5) and practically eliminates eIF1 from the MFC in vivo (Fig. 6). Despite numerous attempts, however, the results were not satisfactorily conclusive because they suffered from huge experimental error. We think the most probable reason for this is the extremely long doubling time of the large internal deletion of this mutant that makes it very hard to work with when compared with other, not so slow growing mutants. Nevertheless, our in vitro and in vivo binding as well as hc eIF1 suppression data strongly suggest that GAP⁸⁵ also causes the Sui⁻ phenotype by weakening the eIF1 binding to PICs in a manner partially suppressible by hc eIF1.

The class “c” WW mutant, which still falls in the original Box12 segment, imparts only the Gcd⁻ phenotype in a manner partially suppressible by hc TC (Fig. 4) and as such was expected to reduce mainly the eIF2 amounts on the 40 S ribosome. As shown in Fig. 7C, besides eIF2, we also found decreased amounts of eIF5 but not eIF3 and eIF1. Interestingly and in accord with our binding data showing a significant loss of eIF2 but not eIF5 from the MFC in vivo (Fig. 6), 40 S ribosome binding of eIF2 was more severely impaired than that of eIF5. This is significant because in all other Gcd⁻ mutants examined in this work eIF5 association with PICs was always more impaired than that of eIF2. These results seem to indicate that the WW mutant impairs eIF2 recruitment by a different mechanism than via the destabilization of eIF5 binding to the c/Nip1-NTD in the MFC, which is the only way that we have seen so far (see “Discussion” for further details). These results also suggest that the “PW” but not the “WW” part of the WWPW mutant is responsible for the weakened eIF3 binding to PICs in vivo.

Finally, another representative class c mutant, Box15-2, reduced binding of eIF3 and eIF5 with 40 S species (Fig. 7D, right panel), indicating that this part of the NTD of c/Nip1 also is required for stable incorporation of eIF3 and eIF5 into the PICs. Importantly, consistent with the Gcd⁻ only phenotype of this mutant, 40 S subunit-associated amounts of eIF2 (TC) were also significantly reduced in contrast to eIF1, the binding of which was affected only modestly.