Regulatory elements in eIF1A control the fidelity of start codon selection by modulating tRNA_i^Met binding to the ribosome

Abstract

eIF1A is the eukaryotic ortholog of bacterial translation initiation factor IF1, but contains a helical domain and long unstructured N-terminal tail (NTT) and C-terminal tail (CTT) absent in IF1. Here, we identify elements in these accessory regions of eIF1A with dual functions in binding methionyl initiator tRNA (Met-tRNA_i^Met) to the ribosome and in selecting AUG codons. A pair of repeats in the eIF1A CTT, dubbed Scanning Enhancer 1 (SE₁) and SE₂, was found to stimulate recruitment of Met-tRNA_i^Met in the ternary complex (TC) with eIF2·GTP and also to block initiation at UUG codons. In contrast, the NTT and segments of the helical domain are required for the elevated UUG initiation occurring in SE mutants, and both regions also impede TC recruitment. Remarkably, mutations in these latter elements, dubbed scanning inhibitors SI₁ and SI₂, reverse the defects in TC loading and UUG initiation conferred by SE substitutions, showing that the dual functions of SE elements in TC binding and UUG suppression are mechanistically linked. It appears that SE elements enhance TC binding in a conformation conducive to scanning but incompatible with initiation, whereas SI elements destabilize this conformation to enable full accommodation of Met-tRNA_i^Met in the P site for AUG selection.

Identification of the translation initiation codon in eukaryotes typically occurs by a scanning mechanism in which the small (40S) ribosomal subunit recruits the methionyl initiator tRNA (Met-tRNA_i^Met) in a ternary complex (TC) with the GTP-bound form of eIF2 to form the 43S preinitiation complex (PIC). The 43S PIC then binds to the mRNA near the m⁷G-capped 5′ end and scans the leader, using complementarity with the anticodon of the initiator as a key means of identifying the AUG start codon (Pestova et al. 2007). The mechanisms involved in ribosomal scanning and in distinguishing AUG from non-AUG triplets by the scanning PIC are not fully understood.

The factors eIF1 and eIF1A have been shown to stimulate scanning and assembly of a stable 48S PIC at AUG, and eIF1 also blocks recognition of near-cognate triplets, in a reconstituted mammalian system (Pestova and Kolupaeva 2002). eIF1 appears to act with eIF1A to promote an open, scanning-conducive conformation of the PIC (Lomakin et al. 2000; Fekete et al. 2007; Passmore et al. 2007), and it impedes GTP hydrolysis by the TC in the absence of perfect base-pairing between the P-site codon and anticodon of Met-tRNA_i^Met (Unbehaun et al. 2004). Consistent with this, recognition of AUG elicits dissociation of eIF1 from the 40S subunit and accelerates P_i release from eIF2·GDP·P_i in a reconstituted yeast system (Algire et al. 2005; Maag et al. 2005).

The mechanisms of scanning and AUG selection are being dissected with genetic tools in budding yeast. Substitutions in the three subunits of eIF2, eIF5, and eIF1 were described that increase the frequency of initiation at the UUG start codon of his4-301 mRNA, restoring the ability to synthesize histidine (His⁺ phenotype) (Yoon and Donahue 1992; Donahue 2000). A subset of such Sui⁻ (Suppressor of initiation codon mutant) substitutions affecting eIF1 (encoded by SUI1) appear to act simply by accelerating eIF1 dissociation from the PIC (Cheung et al. 2007). Conversely, overexpression of wild-type eIF1 suppresses the Sui⁻ phenotypes of the SUI5 and SUI3-2 substitutions in eIF5 and eIF2β, respectively (Valasek et al. 2004; Fekete et al. 2007), conferring the Ssu⁻ (Suppression of Sui⁻) phenotype. These findings support the notion that eIF1 is a “gatekeeper” that impedes start codon selection and whose inhibitory functions are eliminated at AUG codons, at least partly, by its dissociation from the 40S subunit. Presumably, overexpression of eIF1 prevents its release from the 40S subunit, which would otherwise occur at a higher frequency at UUGs in Sui⁻ mutants, and thereby allows scanning to continue downstream.

The eIF1A contains an oligonucleotide/oligosaccharide-binding (OB) fold domain related to that present in bacterial translation initiation factor IF1, but additionally contains a helical domain comprised of helix α2 and a 3₁₀ helix connected by a short linker region (Lαα), plus structured N or C strands that pack against different surfaces of α2. eIF1A also contains long unstructured N-terminal tail (NTT) or C-terminal tail (CTT) (Fig. 1A; Battiste et al. 2000). The original genetic selections that yielded Sui⁻ mutations affecting eIF1, eIF5, or subunits of eIF2 failed to identify Sui⁻ mutations in TIF11 encoding yeast eIF1A (Donahue 2000). However, we found previously that removing the unstructured CTT and the C strand, Lαα, and 3₁₀ elements of the helical domain by the C-terminal truncation Δ108–153 (ΔC) confers a His⁺/Sui⁻ phenotype in the his4-301 background. The ΔC truncation also impaired the ability of eIF1A to promote scanning in a reconstituted mammalian system, leading us to propose that the C-terminal region of eIF1A acts to enhance scanning at non-AUG codons, such that its impairment by ΔC increases UUG initiation. Remarkably, alanine substitutions of residues 17–21 (17–21) in the unstructured NTT confer an Ssu⁻ phenotype, leading to the proposal that the NTT acts to inhibit scanning and promote start codon recognition, such that its inactivation by 17–21 suppresses UUG initiation in Sui⁻ mutants (Fekete et al. 2005, 2007). Thus, the NTT and C-terminal region of eIF1A appeared to have opposite effects on scanning and start codon recognition.

Figure 1.

Mapping the SE elements in the CTT of eIF1A. (A) Schematic showing the domains in eIF1A, indicating the residue number that begins each domain (above), and locations of the SEs in the CTT sequence (below). (Constructs a–p) Schematics indicating amino acids missing (red dashes) in different TIF11 alleles. (Right) Relative growth, on a scale of 0–10, of his4-301 tif11Δ strains harboring the indicated TIF11 alleles on LEU2 plasmids under the conditions described in B. (B–D) Slg⁻ and His⁺/Sui⁻ phenotypes of strains harboring selected TIF11 alleles, described in A, were determined by spotting serial 10-fold dilutions on synthetic complete medium lacking leucine (SC-L) supplemented with 0.3 mM His (+His) or 0.0003 mM His (−His) and incubating for 3 d (+His) or 6 d (−His) at 30°C.

Interestingly, the ΔC truncation, or double-Ala substitution of eIF1A CTT residues Phe-131 and Phe-133, also impairs loading of the TC on 40S subunits in PIC assembly. A well-established manifestation of this defect is the perturbation of translational control of GCN4 mRNA by four short upstream ORFs (uORFs). Ribosomes that have translated the 5′-most uORF in GCN4 mRNA (uORF1) and resumed scanning can bypass the remaining three uORFs (uORF2–uORF4) and reinitiate at the GCN4 AUG when the levels of TC are reduced by eIF2α phosphorylation by protein kinase Gcn2 in amino acid-starved cells. The reduced TC concentration enables a fraction of the reinitiating 40S subunits to rebind TC only after bypassing uORF2–uORF4, but before reaching the GCN4 start codon. A similar shift in reinitiation from uORF2–uORF4 to the GCN4 ORF also occurs in eIF mutants where the rate of TC loading on 40S subunits is impaired, constitutively derepressing GCN4 translation independently of Gcn2 (the Gcd⁻ phenotype) (Hinnebusch 2005). The ΔC and the F131A,F133A mutations confer Gcd⁻ phenotypes by this mechanism, and also reduce the rate of TC loading on 40S subunits in the reconstituted yeast system (Olsen et al. 2003; Fekete et al. 2005, 2007). Thus, the eIF1A CTT appears to have dual functions in promoting TC binding and suppressing UUG initiation, but it was unclear whether these two functions are related.

While the F131A,F133A substitution in eIF1A increases the UUG:AUG initiation ratio measured using HIS4-lacZ reporters differing in these start codons, it does not confer the His⁺/Sui⁻ phenotype observed for the ΔC truncation, suggesting that other residues removed by ΔC besides F131,F133 act to enhance scanning and block initiation at UUG codons (Fekete et al. 2005). Hence, we set out to identify all residues in eIF1A that participate in this key initiation function. In the process, we discovered that two loosely conserved ∼10-residue repeats in the CTT, dubbed scanning enhancers (SEs), are the critical elements that both inhibit UUG initiation and promote TC binding to the 40S subunit, and we provide genetic and biochemical evidence that these activities are mechanistically linked. We further demonstrate that segments of the helical domain of eIF1A function with the NTT to impede scanning and promote start codon recognition and also negatively regulate TC binding. Our results support a model in which SE elements stabilize a mode of TC binding that is conducive with scanning but incompatible with initiation, and thereby block selection of near-cognate codons, whereas scanning inhibitor (SI) elements antagonize this mode of TC binding to enable AUG recognition.

Results

Identification of partially redundant SE elements in the eIF1A CTT

The F131A,F133A mutation in the eIF1A CTT increases the UUG:AUG initiation ratio, but unlike the ΔC truncation, does not confer a His⁺/Sui⁻ phenotype in his4-301 cells (Fekete et al. 2005, 2007). To identify the other CTT residues responsible for the strong Sui⁻ phenotype of Δ108–153, we first constructed the set of nested deletions shown in Figure 1A (constructs b–f) with a common end point at the predicted N-terminal residue of the unstructured CTT (Asp-119) (Battiste et al. 2000). The TIF11 alleles on a LEU2 plasmid were introduced into a his4-301 tif11Δ ura3 strain harboring wild-type TIF11 on a URA3 plasmid, and the latter TIF11⁺ plasmid was evicted by counterselection on medium containing 5-fluoroorotic acid (5-FOA) (Boeke et al. 1987). Western analysis of whole-cell extracts (WCEs) of these and other mutants described below showed that none of their phenotypes could be attributed to reduced eIF1A expression (Supplemental Fig. S1). The deletion mutants were tested for His⁺/Sui⁻ phenotypes by examining cell growth on (−His) medium containing only 0.1% of the amount of histidine added to +His medium to fully supplement His⁻ auxotrophs. Only the largest deletion (Δ119–133), which removed F131 and F133, produced a His⁺ phenotype while also conferring a strong slow-growth (Slg⁻) phenotype on +His medium (Fig. 1A [constructs a–f], B [lanes a,e,f]). Given that substituting F131 and F133 alone did not have a His⁺ phenotype (Fekete et al. 2007), we surmised that residues between 119 and 130 are functionally redundant with F131 and F133. We provisionally dubbed this hypothetical element Scanning Enhancer 1 (SE₁) and that containing F131,F133 SE₂, and hypothesized that a strong Sui⁻ phenotype results only when SE₁ and SE₂ are impaired simultaneously.

To support the idea that SE₁ is functionally redundant with SE₂, we constructed another set of deletions with a common end point at F133, and thus all lacking F131,F133 of SE₂ (Fig. 1A, constructs g–i). Examination of these mutants revealed that removing residues from F133 to S125 was required to evoke His⁺ and Slg⁻ phenotypes (Fig. 1B [lanes i vs. h], summary in 1A [constructs g–i]). Comparing constructs i and h in Figure 1A suggested that the C-terminal boundary of SE₁ lies between S125 and E127. To map the N-terminal boundary of SE₁, we deleted one residue at a time beginning with D119 from construct Δ128–133, which lacks F131, F133 of SE₂ (Fig. 1A, constructs j–l). Moderate His⁺ and Slg⁻ phenotypes appeared only with removal of F121 (Fig. 1C, lanes l vs. k). Together, these results suggested that SE₁ maps between F121 and E127 (Fig. 1A, schematic).

The aforementioned results for constructs e and f in Figure 1A suggested that F131,F133 occupy the N terminus of SE₂. To map the C terminus of SE₂, we deleted residues from the extreme C terminus of eIF1A in a construct where SE₁ is removed by Δ119–130 (Fig. 1A, constructs m–p). Moderate His⁺ and Slg⁻ phenotypes appeared only with the deletion of residues 134–136 in construct p (Fig. 1D, lanes p vs. o), thus suggesting that SE₂ extends from F131 to A136 (Fig. 1A, schematic).

To test directly the conclusion that SE₁ and SE₂ have overlapping functions in blocking UUG initiation, we constructed complete, multiple-Ala substitutions of SE₁ or SE₂. The SE₁* mutation conferred only a slight His⁺ phenotype (Fig. 2B, lanes a,b) and a correspondingly small increase in the UUG:AUG initiation ratio measured using HIS4-lacZ reporters (Fig. 2C, bars a,b). The SE₂* mutation, in contrast, conferred Slg⁻, a stronger His⁺ phenotype (Fig. 2B, lanes b,c), and a much larger increase in the UUG:AUG initiation ratio (Fig. 2C, bars b,c). Strikingly, the SE₁*,SE₂* double mutation is lethal, preventing eviction of the TIF11⁺ plasmid (data not shown). However, combining SE₁* with an incomplete substitution of SE₂ in which F131 is retained (mutation SE₂* + F₁₃₁) yields a viable mutant with strong Slg⁻ and His⁺ phenotypes (Fig. 2B, lane d) and an even larger UUG:AUG initiation ratio than seen for SE₂* alone (Fig. 2C, bars d vs. c). By assaying a pair of luciferase (LUC) reporters with UUG or AUG start codons, we confirmed that SE₁*,SE₂* + F₁₃₁ provokes a larger increase in the UUG:AUG initiation ratio than does SE₂* or SE₁* (Fig. 2E, bars a–d). These findings support the idea that SE₁ and SE₂ are partially redundant elements that block initiation at UUG, and suggest that SE₂ is more crucial than SE₁ for this function.

Figure 2.

SE₁ and SE₂ elements functionally overlap in blocking UUG initiation. (A) Schematic of eIF1A showing sequence similarity between SE₁ and SE₂. Residues 121–127 were substituted with Ala in mutant SE₁*, and residues 131–135 were similarly substituted in SE₂*. (B) Slg⁻ and His⁺/Sui⁻ phenotypes of his4-301 strains containing the indicated TIF11 alleles determined as described in Figure 1B. (C,D) Strains from B harboring HIS4-lacZ reporter plasmids with an AUG (p367) or UUG (p391) start codon were cultured in SC lacking leucine and uracil (SC-LU) at 30°C, and β-galactosidase activities (nanomoles of o-nitrophenyl β-D-galactopyranoside cleaved per minute per milligram) were measured in WCEs. (E) his4-301 strains with the indicated TIF11 alleles and harboring the dual luciferase reporter plasmid pRaugFFuug, containing LUC_Renilla and LUC_firefly coding sequences with AUG or UUG start codons, and driven by the ADH1 or GPD promoters, respectively, were cultured in SC-LU, and luminescence was measured in WCEs. The ratio of luminescence, in relative light units, for the UUG to AUG reporter was calculated and plotted. Error bars in C–E give the standard errors of the mean. Differences in bracketed mean values were judged to be significant by the Student's t-test, with P < 0.001 (***).

Noting that SE₁ and SE₂ comprise a loosely conserved repeat of nine to 10 residues, each containing a pair of Phe residues (Fig. 2A), we sought to demonstrate that the Phe pairs are the critical residues in SE₁ and SE₂ blocking UUG recognition. First, we generated constructs combining a complete substitution of one SE element with an incomplete substitution of the other SE that retains both Phe residues and compared them with construct SE₁*,SE₂* + F₁₃₁ containing only one Phe residue. Both constructs retaining two Phe residues (Fig. 2B, lanes e,f) produce weaker Slg⁻ and His⁺ phenotypes (Fig. 2B) and lower UUG:AUG ratios (Fig. 2C) compared with that with only a single Phe (Fig. 2C, bar d). Furthermore, a construct (Fig. 2C, bar g) in which all of the residues in the two SEs are substituted except for the four Phe residues produces no Slg⁻ or His⁺ phenotype (Fig. 2B, lanes g vs. a) and only a slight increase in the UUG:AUG ratio compared with wild type (Fig. 2C, bars g vs. a). These results suggest that the number of Phe residues is the critical determinant of the Slg⁻ and Sui⁻ phenotypes of SE mutations.

Finally, we compared substitutions of only the two Phe residues in SE₁ versus those in SE₂. F_131,133A provokes a stronger His⁺ phenotype (Figs. 2B, lanes h,i) and a larger increase in UUG:AUG ratio (Fig. 2D) compared with F_121,123A. Although neither double-Phe substitution confers a Slg⁻ phenotype, the quadruple-mutant F_{121,123,131,133}A is lethal (data not shown). Thus, F121,F123 are less critical than F131,133, but become essential for blocking UUG initiation in the absence of F131,F133. (The fact that the F_131,133A mutant is His⁺ and Slg⁺ does not contradict our previous findings, as we had analyzed a Flag-tagged version of this allele. The Flag tag reduces the His⁺/Sui⁻ phenotypes and exacerbates the growth defects of various TIF11 mutations [Fekete et al. 2007].) Together, the results in Figure 2 indicate that the phenylalanines are the critical residues in SE₁ and SE₂, which act redundantly to suppress UUG initiation and support cell growth.

Evidence that elimination of both SEs is lethal owing to an intolerable defect in scanning or non-AUG initiation

It is possible that precisely substituting both SEs is lethal because of an intolerable increase in non-AUG initiation. To test this idea, we asked whether the lethality could be overcome by the Ssu⁻ mutation 17–21 in the NTT of eIF1A or by overexpressing eIF1, both of which reduce UUG initiation in other Sui⁻ mutants. Indeed, 17–21 suppresses the lethality of SE₁*,SE₂*, although the resulting mutant retains Slg⁻ and His⁺/Sui⁻ phenotypes (Fig. 3A, +His and −His, segments c vs. a) and a high UUG:AUG ratio of 0.5 ± 0.05. Importantly, additionally overexpressing eIF1 from a high-copy (hc) SUI1 plasmid improved the growth rate and nearly eliminated the His⁺/Sui⁻ phenotype of the SE₁*,SE₂*,17-21 strain (Fig. 3A, +His and −His, segments d vs. c). Similarly, introducing either 17–21, hc SUI1, or both into the viable mutant SE₁*,SE₂* + F₁₃₁ suppressed its His⁺/Sui⁻ phenotype (Fig. 3A, −His, segments f–h vs. e) and (for 17–21) diminished the elevated UUG:AUG ratio in this viable Sui⁻ mutant (Fig. 2E, bars e vs. d).

Figure 3.

SI element substitutions in eIF1A, or eIF1 overexpression, suppresses the Sui⁻ phenotypes of SE substitutions. (A) his4-301 strains with the indicated TIF11 alleles plus hc SUI1 plasmid YEpW-SUI1 or empty TRP1 vector were streaked on SC lacking leucine and tryptophan (SC-LW) with 0.3 mM His (+His) or 0.0003 mM His (−His) and incubated for 3 d (+His) or 6 d (−His). (B) his4-301 strains containing the indicated TIF11 alleles or empty LEU2 vector and harboring TIF11⁺ on a URA3 plasmid (lanes a–f), and also hc SUI1 or empty TRP1 vector (lanes e,f), were replica-plated to either SC-LU (lanes a–d) or SC-LUW (lanes e,f), both containing 0.3 mM His (+His) or 0.0003 mM His (−His), and to either SC-L (lanes a–d) or SC-LW (lanes e,f), both containing 5.2 mM 5-FOA. Plates were incubated for 3 d (+His), 6 d (−His), or 8 d (5-FOA). (C) His⁺/Sui⁻ phenotypes of his4-301 strains containing the indicated TIF11 alleles plus hc SUI1 or empty vector were determined as in B. (D) his4-301 strains with the indicated TIF11 alleles, TIF11⁺ on a TRP1 vector, and the AUG or UUG HIS4-lacZ reporters were analyzed as in Figure 2C except that cells were cultured in SC-LUW. (E) his4-301 strains with the indicated TIF11 alleles, hc SUI1 or empty TRP1 vector, and the AUG or UUG HIS4-lacZ reporters were analyzed as in D. (F) Schematic of eIF1A and the extent of C-terminal truncations in the indicated TIF11 alleles, with a summary of phenotypes, as in Figure 1A. (G) Slg⁻ and His⁺/Sui⁻ phenotypes of his4-301 strains containing the indicated TIF11 alleles were determined as in Figure 1B. (H) his4-301 strains with the indicated TIF11 alleles and the AUG or UUG HIS4-lacZ reporters were analyzed as in Figure 2C.

Consistent with the above findings, Δ117–153, which removes both SE₁ and SE₂, is also lethal, and this lethality is suppressed by 17–21 (Fig. 3B, 5-FOA, lanes c vs. d). Interestingly, Δ117–153 confers a His⁺/Sui⁻ phenotype in viable cells also harboring TIF11⁺, and this dominant His⁺/Sui⁻ phenotype is suppressed by 17–21 or hc SUI1 (Fig. 3B, −His, lanes c,d and e,f). As shown above for SE₁*,SE₂*,17–21, introducing hc SUI1 into the viable Δ117–153,17–21 mutant eliminates its residual His⁺/Sui⁻ phenotype (Fig. 3C, lanes c vs. d).

The aforementioned results indicate that the lethality of eliminating both SEs involves an extreme defect in non-AUG initiation that is lessened by an Ssu⁻ mutation in eIF1A or overexpressing eIF1. The dominant Sui⁻ phenotype of Δ117–153 further indicates that this defective protein can compete with wild-type eIF1A for incorporation into PICs, but then fails to block UUG initiation effectively. Given the proposed function of eIF1 in promoting the open, scanning conformation of the PIC, and the ability of eIF1 overexpression or the 17–21 mutation to suppress the lethality and Sui⁻ phenotypes caused by eliminating both SE elements, we infer that the SE elements promote the open, scanning conformation of the PIC (hence, their designation as SEs). In contrast, the eIF1A NTT inhibits this open conformation and is regarded as an SI element, which we dubbed SI₁.

Identification of a second SI element in the eIF1A helical domain

Our previous finding that deleting residues 108–153 from eIF1A (ΔC) confers a Sui⁻ phenotype (Fekete et al. 2005) is consistent with the elimination of both SE elements, but seems at odds with the lethality observed here for Δ117–153 and SE₁*,SE₂*, which likewise eliminate both SEs . Because the lethality of the latter mutations is suppressed by the 17–21 mutation in SI₁, we reasoned that Δ108–153 is viable because it deletes a second SI element (provisionally dubbed SI₂) that is located just upstream of the SEs and would be left intact by the smaller truncation Δ117–153. Supporting this interpretation, extending the deletion in Δ117–153 by only the single codon for Asn-116 (producing Δ116–153) suppressed the lethality of Δ117–153 (data not shown). Furthermore, Δ116–153 confers a smaller increase in the UUG:AUG initiation ratio than does Δ117–153 when these alleles are compared in situations where Δ117–153 is not lethal, either in the presence of TIF11⁺ (Fig. 3D) or hc SUI1 (Fig. 3E). Δ116–153 also confers weaker Slg⁻ and His⁺/Sui⁻ phenotypes than does Δ117–153 in the presence of hc SUI1 (Supplemental Fig. S2A, lanes d vs. e). Thus, it appears that Asn-116 belongs to the putative SI₂ and its elimination by Δ116–153 suppresses the lethal, hyper-Sui⁻ phenotype of removing both SEs in Δ117–153.

To determine whether elimination of N116 fully inactivated SI₂, we examined the Sui⁻ phenotypes of constructs that extend the deletion of C-terminal domain (CTD) residues further upstream to also remove P110, G107, or Q106 (Fig. 3F, constructs c–f). Compared with Δ116–153, these larger deletions improved the growth rate on complete medium and progressively reduced the Sui⁻ phenotype (Fig. 3G, +His and −His, lanes d–f vs. c) and the UUG:AUG initiation ratio (Fig. 3H, bars c–e vs. b). (The weaker His⁺ phenotype observed for Δ116–153 vs. Δ110–153 likely reflects the relatively poor growth of the Δ116–153 mutant evident on +His medium.) Thus, extending the deletion of CTD residues from N116 to Q106, removing the entire C strand, 3₁₀ helix, and linker Lαα (Fig. 3F), constitutes a step-wise reduction in SI₂ function.

To confirm this last conclusion and also map the N-terminal boundary of SI₂, we made nested Ala substitutions from residue 106 to 109 (the last residue of α2 and Lαα linker) in the lethal Δ117–153 truncation (Fig. 4A, “A” substitutions in constructs c–f). We reasoned that substitutions encroaching into the N terminus of SI₂ should confer an Ssu⁻ phenotype and suppress the lethal, hyper-Sui⁻ phenotype of Δ117–153. Indeed, substituting both Q106 and G107, but not Q106 alone, suppressed the lethality of Δ117–153, and the more extensive substitutions that include residues 108 and 109 conferred even higher growth rates (Supplemental Fig. S2B, 5-FOA, lanes c–f vs. b; results summarized in Fig. 4A, +His). Importantly, the 106–109 substitution (106–109Ala) also suppressed the lethality of the hyper-Sui⁻ mutation SE₁*,SE₂* (data not shown), and the residual His⁺/Sui⁻ phenotype of the resulting viable SE₁*,SE₂*,106–109Ala mutant was eliminated by extending the Ala substitutions to encompass residues 106–116 in mutant SE₁*,SE₂*,106–116Ala (Fig. 4B, −His, sections c,d). Mutation 106–109Ala also reduced the Slg⁻ and Sui⁻/His⁺ phenotypes of the viable mutant SE₁*,SE₂* + F₁₃₁ (Fig. 4B, sections h vs. g). Together, the results suggest that SI₂ spans residues 107–117 and encompasses Lαα, the 3₁₀ helix, and most of the structured C strand in the eIF1A CTD.

Figure 4.

Substitutions in region SI₂, which packs against α2, confer an Ssu⁻ phenotype. (A) Mapping the N-terminal boundary of SI₂. Schematic of eIF1A and the residues lacking (red dashes) or substituted with alanines (As) in the indicated TIF11 alleles, with a summary of phenotypes, as in Figure 1A. (B) Slg⁻ and His⁺/Sui⁻ phenotypes of his4-301 strains containing the indicated TIF11 alleles were determined as in Figure 3A, except the cells were streaked on SC-L medium, and are summarized on the right. (C,D) Positions of conserved residues E₁₀₈P₁₁₀E₁₁₁I₁₁₅N₁₁₆ in human eIF1A and their probable contacts with the indicated residues in helix α2 helix (C), and of N strand residues L₂₆Y₂₈K₂₉ and their predicted contacts with α2 and Lαα residues (D). The coordinates of human eIF1A (PDB:1D7Q) were used to generate structural models using PyMOL software, showing selected side chains in stick representation.

Because all of the components of SI₂ belong to the helical domain (Fig. 4C), we asked whether altering residues in the Lαα, 3₁₀ helix, or C strand that interacts directly with α2 (Battiste et al. 2000) would also confer an Ssu⁻ phenotype by disrupting SI₂ function. We first examined the effects of Ala substitutions in five such C strand residues: Glu108, Pro110, Glu111, Ile115, and Asn116 (Fig. 4C). This mutation (EPEIN-Ala) suppressed the lethality of the hyper-Sui⁻ allele Δ117–153 (Supplemental Fig. S2B, 5-FOA, lanes b vs. g) as described above for Ssu⁻ mutation 106–109Ala. Because the structured N strand of the N-terminal domain (NTD) (residues 26–33) also packs against α2, we then asked whether substituting N strand residues Leu26, Tyr28, and Lys29 that are predicted to contact residues in α2 or Lαα (Fig. 4D) would likewise confer Ssu⁻ phenotypes. Indeed, this triple-Ala substitution (LYKAla) suppressed the lethality of Δ117–153 (Supplemental Fig. S2B, 5-FOA, lanes b vs. h) and diminished the elevated UUG:AUG initiation ratio conferred by Δ116–153, closely resembling the Ssu⁻ mutation 17–21 (Fig. 3H, bars f,g vs. b). Consistent with these findings, the N and C strand mutations LYKAla and 106–109Ala decreased the Slg⁻ and His⁺/Sui⁻ phenotypes of the viable mutant SE₁*,SE₂* + F₁₃₁ (Fig. 5A, sections a–c vs. d), and, along with 106–116Ala, also reduced the UUG:AUG ratio in SE₁*,SE₂* + F₁₃₁ cells (Fig. 5B, bars f–h vs. e). These results indicate that proper packing of both N and C strands against α2 is required for the function of SI₂ in supporting UUG initiation in Sui⁻ mutants of eIF1A.

Figure 5.

SI substitutions suppress the Sui⁻ phenotype of SUI3-2. (A) Slg⁻ and His⁺/Sui⁻ phenotypes of his4-301 strains containing the indicated TIF11 alleles were determined as in Figure 4B. (B) his4-301 strains with the indicated TIF11 alleles plus the dual luciferase reporter pRaugFFuug were assayed as described in Figure 2E. (C) Slg⁻ and His⁺/Sui⁻ phenotypes of his4-301 strains with the indicated TIF11 alleles and containing either plasmid pRSSU13-S264Y-W harboring SUI3-2 (lanes 2–7) or empty vector (lane 1) were determined as in Figure 1B, except that the cells were spotted on SC-LW medium. (D) Strains described in C harboring the AUG or UUG HIS4-lacZ reporters were analyzed as in Figure 2C. (E) Predicted interaction of eIF1A residues D₉₈E₉₉A₁₀₀R₁₀₁ with C and N strand residues is depicted as in Figure 4C.

We asked next whether mutations affecting SI₂ can suppress the Sui⁻ phenotype of SUI3-2, encoding the S264Y substitution in eIF2β. Remarkably, 106–116Ala and LYKAla both diminish the Sui⁻/His⁺ phenotype of SUI3-2 cells (Fig. 5C; data not shown) and reduce the elevated UUG:AUG initiation ratio conferred by SUI3-2 (Fig. 5D), mimicking the Ssu⁻ mutation 17–21 in SI₁ in both respects. Interestingly, the DEAR-AASA mutation described previously (Fekete et al. 2007), which substitutes Asp98, Glu99, and Arg101 of α2 itself and eliminates contacts with C or N strand residues (Fig. 5E), also confers an Ssu⁻ phenotype in SUI3-2 cells (Fig. 5C,D). Thus, the integrity of the entire helical domain is required to support the elevated UUG initiation rate in SUI3-2 cells. Finally, eliminating the entire NTT by the ΔN mutation confers a stronger Ssu⁻ phenotype compared with 17–21 (Fig. 5D), indicating that 17–21 only partially impairs SI₁.

Gcd⁻ phenotypes of Sui⁻ mutations in SE₁SE₂ are suppressed by Ssu⁻ mutations in SI₂

We showed previously that the ΔC and F_131,133A mutations impair TC binding to 40S subunits. When introduced into Flag-tagged eIF1A, these substitutions derepress translation of GCN4 mRNA independently of eIF2α phosphorylation by GCN2. This Gcd⁻ phenotype was attributed to a reduced rate of TC loading on 40S subunits that have translated uORF1 and resumed scanning, allowing them to bypass uORF2–uORF4 and reinitiate at GCN4 without any decrease in TC formation (Olsen et al. 2003; Fekete et al. 2005, 2007). Consistent with these findings, the SE₂* mutation (which eliminates F131,F133) confers a pronounced Gcd⁻ phenotype, derepressing by ∼10-fold the GCN4-lacZ reporter containing all four uORFs in nonstarved GCN2 cells (Fig. 6A, Nonstarvation, bars c vs. a). The SE₁* mutation confers a smaller derepression of GCN4-lacZ, and adding to it the Ala substitutions of all SE₂ residues except F131 (mutant SE₁*,SE₂*+F₁₃₁) evokes a stronger derepression than does SE₁* alone (Fig. 6A, Nonstarvation, bars d vs. b). None of these mutations increases expression of a GCN4-lacZ construct lacking all four uORFs in nonstarvation conditions (Supplemental Fig. S3A), confirming that they diminish the translational repression imposed by the uORFs. Importantly, co-overexpressing all three eIF2 subunits and tRNA_i^Met, the components of TC, from an hc plasmid (hc TC) reduces the derepression evoked by the SE₁*, SE₂*, and SE₁*,SE₂* + F₁₃₁ mutations (Fig. 6C). This supports the idea that their Gcd⁻ phenotypes arise from slower TC loading on reinitiating 40S subunits scanning downstream from uORF1, which is overcome by mass action at higher concentrations of TC.

Figure 6.

In vivo and in vitro TC loading defects of SE substitutions are suppressed by SI substitutions. (A) GCN2 strains containing the indicated TIF11 alleles and GCN4-lacZ reporter plasmid p180, depicted at the top, were cultured in repressing (Nonstarvation) medium (SC-LU) or derepressing (Starvation) medium (SC-LU lacking isoleucine and valine and supplemented with 0.5 μg/mL sulfometuron), and β-galactosidase activities were assayed in WCEs. (B) GCN2 strains containing the indicated TIF11 alleles were analyzed as in A under nonstarvation conditions. (C) GCN2 strains harboring the indicated TIF11 alleles, GCN4-lacZ reporter plasmid p180, and either empty vector, hc SUI1, or hc TC plasmid p4385 were assayed under repressing conditions as in A, except that cells were grown in SC-LUW. (D) GCN2 strains harboring the indicated TIF11 alleles plus the AUG or UUG HIS4-lacZ reporters and either empty vector, hc SUI1, or hc TC were analyzed as in Figure 3D. (E) GCN2 strains harboring the indicated TIF11 alleles, GCN4-lacZ reporter plasmid p180, and either empty vector or SUI3-2 were assayed under repressing conditions as in C. (F) Binding of TC to 40S subunits as a function of the concentration of 40S subunits (K_d), or time (k_obs), measured as the fraction of [³⁵S]-Met-tRNA_i^Met associated with 40S subunits in a native gel assay containing saturating eIF1 (1 μM), model mRNA(AUG), and wild-type or mutant eIF1A proteins at 1 μM. Errors are mean deviations of at least three measurements. (G) Effects of eIF1A substitutions on the rate of ribosomal subunit joining. Kinetics of subunit joining measured using purified 40S and 60S subunits at 100 nM and 200 nM, respectively, and wild-type or mutant eIF1A proteins at 1 μM by monitoring the change in intensity of light scattering over time. Kinetics were biphasic for wild type, 106–116Ala, and 17–21, but monophasic for the other three eIF1A mutants. (k₁ and k₂) Rate constants for the first and second phases, respectively; (α1/α2) ratio of amplitudes of the first to second phases; (na) not applicable owing to monophasic reactions. Errors are mean deviations of at least three measurements.

The Ssu⁻ mutations 106–109Ala and 106–116Ala in SI₂ do not provoke Gcd⁻ phenotypes; rather, 106–116Ala has the opposite effect of reducing GCN4-lacZ derepression under starvation conditions, indicating a Gcn⁻ phenotype (Fig. 6A, Starvation, bars e,f vs. a). Ssu⁻ mutations in SI₁, including ΔN, 17–21, and G21D, confer comparable Gcn⁻ phenotypes (Supplemental Fig. S3B). Importantly, the Ssu⁻ mutations 106–109Ala, 106–116Ala, 17–21, and LYKAla all diminish the derepression of GCN4-lacZ expression conferred by SE₁*,SE₂*+F₁₃₁, with 106–116Ala having the greatest effect (Fig. 6B, bars c–f vs. b). The Sui⁻ mutation in eIF2β SUI3-2 has a Gcd⁻ phenotype (Williams et al. 1989), and, remarkably, the Ssu⁻ mutations ΔN, 17–21 and 106–116Ala all greatly suppress the derepression of GCN4-lacZ in SUI3-2 cells (Fig. 6E). Together, these findings indicate that the SE and SI elements have opposite effects on GCN4 translational control. Whereas SE mutations appear to reduce the rate of TC loading and confer Gcd⁻ phenotypes, the Gcn⁻ phenotypes of SI mutations could indicate that they enhance TC binding to reinitiating 40S subunits scanning the leader of GCN4 mRNA.

Substituting SE₁SE₂ impairs TC binding in vitro in a manner suppressed by Ssu⁻ substitutions in SI elements

To test our prediction that SE and SI substitutions have opposite effects on TC loading, we measured their effects on TC binding to 40S subunits in the yeast reconstituted system (Algire et al. 2002; Acker et al. 2007). We first examined the effects of SE and SI substitutions on the equilibrium binding constant of eIF1A for 40S subunits in the presence of saturating concentrations of wild-type eIF1, using a previously established assay (Maag and Lorsch 2003). The results in Supplemental Fig. S4A indicated that only the two mutants harboring the LYKAla substitution displayed an appreciable reduction in eIF1A affinity for 40S subunits. To compensate for these defects in subsequent TC-binding assays, we used eIF1 and eIF1A at concentrations of 1 μM, >80-fold above the K_d values measured for all mutant eIF1A proteins.

To assay TC loading, preformed TC containing [³⁵S]-labeled Met-tRNA_i^Met was incubated with purified 40S subunits, a model mRNA with an AUG start codon, eIF1, and either wild-type or mutant eIF1A, and TC binding to the 40S subunit was monitored by an electrophoretic mobility shift (Kolitz et al. 2009). Importantly, the SE₁*,SE₂* compound substitution markedly decreased the affinity (greater than eightfold increase in K_d) and rate (>35-fold decrease in k_obs) of TC binding to the PIC (Figs. 6F; Supplemental Fig. S4B). The individual SE₁* and SE₂* substitutions also reduced the k_obs for TC loading, but by smaller amounts—less than twofold and twofold, respectively (Fig. 6F). These results fit with the fact that SE₁* and SE₂* produce moderate and strong Gcd⁻ phenotypes, respectively (Fig. 6A), whereas SE₁*,SE₂* is lethal in vivo.

In contrast, the Ssu⁻ substitutions LYKAla, 106–109Ala, and 106–116Ala all increased the rate of TC loading in vitro compared with that seen for wild-type eIF1A (Fig. 6F). Remarkably, all three Ssu⁻ substitutions also greatly reduced the deleterious effects of SE₁*,SE₂* on the affinity and rate of TC loading, restoring the K_d values to <0.5 nM and increasing k_obs by factors of 7–11 compared with the the SE₁*,SE₂* mutant alone (Fig. 6F). The SI₁ substitution 17–21 also strongly suppressed the defective rate of TC binding conferred by SE₁*,SE₂*, although it did not suppress the reduction in TC-binding affinity (Fig. 6F). These results suggest that both SI₁ and SI₂ negatively regulate the rate of TC loading, and that substitutions in these regions compensate for the impaired TC binding conferred by SE substitutions. The fact that SI mutations cosuppress the TC-binding defects in vitro and the Gcd⁻ and Sui⁻ phenotypes in vivo of SE mutations strongly suggests that all of these defects are linked mechanistically to the rate and stability of TC binding.

Substituting the SE elements stabilizes a closed conformation of the 40S subunit competent for subunit joining

Cryo-electron microscopy (cryo-EM) reconstructions of the 40S subunit in the presence or absence of eIF1 and eIF1A demonstrated that the two factors synergistically stabilize the open conformation of the subunit, and biochemical experiments strongly suggested that TC initially binds to this open state (Passmore et al. 2007). Accordingly, we hypothesized that the eIF1A SEs might stimulate the rate of TC loading, at least in part, by maintaining the open conformation of the 40S subunit. To test this possibility, we exploited the observation that eIF1 and eIF1A synergistically impede joining of the 60S with the 40S subunit, consistent with the idea that the open conformation is not receptive to subunit joining and is instead optimized for scanning, whereas subunit joining occurs in the closed conformation following AUG recognition and eIF1 release (Acker et al. 2006). Hence, if mutating the SE elements shifts the equilibrium from the open to closed conformation of the 40S subunit, this should stimulate the rate of 60S joining with the 40S subunit in vitro.

Spontaneous joining of the 40S and 60S subunits follows biphasic kinetics, with the fast and slow phases having roughly equal amplitudes. These two phases likely correspond to different conformational states of the subunit; the fast phase may be 60S joining to the closed state of the 40S subunit and the slow phase may be the conversion of the open to closed state of the 40S subunit (Acker et al. 2006). In the presence of the SE₁*,SE₂* mutant, the kinetics of subunit joining were completely monophasic, with a rate constant (0.05 s⁻¹) similar to that of the fast phase of joining observed for wild-type eIF1A (Fig. 6G; Supplemental Fig. S4C). This is consistent with the idea that substituting the SE elements shifts the conformational equilibrium of the 40S subunit in favor of the closed state, which is competent for subunit joining but not for the initial stage of TC loading. In contrast, the SI substitutions 17–21 and 106–116Ala have the opposite effect of increasing the amplitude of the slow phase of the subunit joining reaction (Fig. 6G), which would be expected if they stabilize the open conformation as a means of enhancing the rate of TC binding. Strikingly, the SI substitutions 17–21 and 106–116Ala also partially suppress the effect of SE₁*,SE₂* by reducing the rate of subunit joining compared with that given by SE₁*,SE₂* alone (Fig. 6G; Supplemental Fig. S4C). The fact that these SI mutations do not fully restore subunit joining to wild-type behavior is consistent with the fact that they only partially suppress the TC loading defects of the SE₁*,SE₂* substitution. These findings strongly suggest that the SE and SI elements regulate the open-to-closed transition of the 40S in opposite directions, which likely contributes to their opposing effects on the rate of TC loading.

Discussion

Previously, we provided evidence that the CTT and NTT of eIF1A differentially regulate the open and closed conformations of the 40S subunit as the means of ensuring stringent selection of AUG as a start codon (Maag et al. 2006; Fekete et al. 2007). Here, we provide additional evidence for this model, but, importantly, we also demonstrate that the SE elements in the CTT and SI elements in the NTT and helical domain regulate scanning and initiation by their differential effects on TC binding to the 40S subunit. By fine-structure mutational analysis, we identified a segment in the unstructured CTT containing loosely conserved 10-amino-acid repeats that act in a partially redundant manner to suppress initiation at the UUG start codon at his4-301. We dubbed these repeats SE₁ and SE₂ because they appear to promote the open conformation of the PIC that facilitates continued scanning when a non-AUG codon occupies the P site. In-depth analysis revealed that these elements contain a pair of Phe residues—F121,F123 and F131,F133, respectively—as critical constituents, and that SE₂ is more important than SE₁ in blocking UUG initiation.

We showed previously that the Δ108–153 (ΔC) truncation of the entire CTT, C strand, 3₁₀ helix, and part of the Lαα linker, produces a viable Sui⁻ mutant (Fekete et al. 2005). Remarkably, we found here that truncations smaller than ΔC that eliminate both SEs without encroaching extensively into the C strand (e.g., Δ117–153) are lethal, as is precisely substituting both SEs by the SE₁*,SE₂* mutation. Because the dominant Sui⁻ phenotypes and lethality of these mutations are suppressed by overexpressing eIF1 (hc SUI1) or the Ssu⁻ mutation 17–21, which also suppress the Sui⁻ phenotypes of mutations in eIF5 (SUI5) and eIF2β (SUI3-2), we can infer that the lethality of removing both SEs involves an intolerably high level of initiation at UUG or other near-cognate start codons that can be mitigated by eIF1 overexpression or NTT substitutions. Although no TIF11 mutations were isolated in the original selections that yielded Sui⁻ mutations (Donahue 2000), our results show that substituting both Phe pairs in the eIF1A SEs increases UUG initiation comparably with the strongest Sui⁻ mutations of eIF1, eIF5, or eIF2. Aligning eIF1A sequences from different species indicates that F121, F131, and F133 are conserved among lower eukaryotes, but only F121 and F131 are conserved in animals (Supplemental Fig. S5). Our results indicate that both F131 and F133 contribute to the function of SE₂ (Fig. 2), but F121 might suffice for SE₁ function.

The fact that Δ117–153 is lethal but the more extensive truncation Δ108–153 (ΔC) is viable was the critical clue that the latter disrupts an element whose elimination compensates for the lethality of deleting both SEs. By systematically testing additional mutations for suppression of the lethality of Δ117–153, we determined that this new element includes the Lαα, 3₁₀ helix, and residues in the structured N and C strands that contact α2 in the helical domain. Similar to the 17–21 mutation in the NTT, disrupting the integrity of the helical domain suppresses Sui⁻ mutations in the SE elements and in eIF2β (SUI3-2), conferring the Ssu⁻ phenotype. Hence, we conclude that α2 and its associated N and C strands comprise a functional unit, dubbed SI₂, that acts in conjunction with SI₁ to arrest scanning and promote start codon selection.

Interestingly, mutations affecting SE₁ and SE₂ also confer strong Gcd⁻ phenotypes that can be suppressed by overexpressing TC, indicating reduced rates of TC loading during reinitiation on GCN4 mRNA. Just as in blocking UUG initiation, SE₂ is more critical than SE₁ in promoting TC loading and repressing GCN4 translation. Our biochemical analysis in the reconstituted system supports this conclusion by revealing reduced rates of TC binding that are more severe for SE₂ versus SE₁ substitutions, and of greatest severity when both elements are lacking. This last result indicates that SE₁ and SE₂ have overlapping functions in TC loading in addition to blocking UUG initiation. The strong correlation between the effects of different SE mutations on the degree of Sui⁻ phenotypes in vivo and the severity of Gcd⁻ phenotypes and TC binding defects in vitro provides evidence that the increased UUG initiation in SE mutants is linked mechanistically to their defective TC binding.

Remarkably, Ssu⁻ mutations in either SI element diminish the Gcd⁻ phenotypes of both SE mutations and the eIF2β Sui⁻ mutation SUI3-2, thus suggesting that the Ssu⁻ mutations compensate for defects in TC binding. Our biochemical data support this conclusion, as defects in TC binding provoked by the SE₁*,SE₂* substitution are partially suppressed by multiple Ssu⁻ substitutions in SI₁ or SI₂, all of which elicit an increased rate of TC binding when present in otherwise wild-type eIF1A. Together, these results provide compelling evidence that the ability of SI mutations to suppress UUG initiation in Sui⁻ mutants involves their ability to compensate for TC loading defects. This leads us to the surprising deduction that the wild-type SI elements act to oppose continued scanning and promote start codon recognition by negatively regulating TC binding.

How might the defect in TC binding provoked by SE mutations lead to increased UUG initiation? We begin by proposing that the SEs stabilize TC binding to the PIC in a conformation that is compatible with scanning but incompatible with initiation. For example, the SEs could bind Met-tRNA_i^Met in a way that prevents it from being fully engaged with the P site—a mode of binding we dub the “P_out” state, which would be associated with the open, scanning conformation of the PIC (Fig. 7A). The SEs could promote the P_out mode of TC binding directly (e.g., by interacting with the anticodon stem–loop [ASL] of the initiator), or indirectly by stabilizing the open conformation of the 40S subunit, to which TC initially binds (Passmore et al. 2007). In either case, entry of AUG into the P site and its perfect complementarity with the anticodon of the initiator would be required to overcome the SEs and enable Met-tRNA_i^Met to be fully accommodated in the P site and trigger downstream steps in the initiation pathway. We dub this second conformation the “P_in” state, and envision that it is characteristic of the closed, scanning-incompatible conformation of the PIC (Fig. 7A). These two states could be identical to the two states that bind TC weakly or stably, respectively, that we detected previously in kinetic studies (Kolitz et al. 2009).

Figure 7.

(A) Model describing the positive and negative effects of the SE and SI elements of eIF1A, respectively, on TC binding in the P_out conformation, which is conducive to scanning, and the second function of the SEs in blocking TC binding in the P_in conformation, which is incompatible with scanning and permissive for initiation. (B) SE inactivation destabilizes P_out, reducing TC loading and conferring the Gcd⁻ phenotype and also enhancing the P_out-to-P_in transition at UUGs to confer the Sui⁻ phenotype. The P_out-to-P_in transition and UUG initiation is further stimulated by loss of the inhibitory effect of the SEs on P_in. (C) SI inactivation stabilizes P_out, promoting TC loading and replacing Gcd⁻ phenotypes with Gcn⁻ phenotypes, and suppressing the P_out-to-P_in transition at UUGs to confer the Ssu⁻ phenotype. (D,E) Hypothetical model depicting the different conformations of initiator tRNA in the P_out (D) and P_in (E) states and the proposed roles of the eIF1A SE elements in stabilizing initiator binding in P_out, where the initiator is not fully accommodated in the P site, and impeding initiator binding in P_in, where the initiator is more fully engaged with the P site. On AUG recognition, the SEs are ejected from the P site to allow greater accommodation of the initiator in the P_in state.

As SE substitutions would impair the ability of eIF1A to stabilize the P_out conformation in which TC initially binds, this would account for their deleterious effect on TC binding to the 40S subunit in vitro and their Gcd⁻ phenotypes in vivo. By destabilizing P_out, SE mutations would also facilitate the P_out-to-P_in transition at UUG codons, which occurs without a perfect codon–anticodon match, and this would help explain their Sui⁻ phenotypes (Fig. 7B). Thus, although SE substitutions reduce the rate of TC loading, once TC is bound to the 40S subunit, the transition from P_out to P_in would occur more frequently at near cognate codons. We found that TC overexpression suppresses the Gcd⁻ phenotypes of SE mutants (Fig. 6C) but does not reduce the elevated UUG:AUG ratio in SE₁*,SE₂* + F₁₃₁ cells (Fig. 6D). These findings are consistent with our model, as increasing TC levels should boost the rate of TC binding to the P_out state, reducing the Gcd⁻ phenotype, but should not mitigate the increased probability of P_out-to-P_in transitions at UUG codons that, in our model, contribute to the Sui⁻ phenotypes of SE mutants.

Our proposal that SE elements stabilize TC binding in a conformation that facilitates scanning but is incompatible with initiation predicts that Met-tRNA_i^Met binds in the P site of the scanning PIC (the P_out state) in a manner distinct from that seen in crystal structures of bacterial 70S·mRNA·tRNA complexes (Berk et al. 2006; Korostelev et al. 2006; Selmer et al. 2006). In fact, this prediction is strongly supported by results of directed hydroxyl radical mapping of the mammalian eIF1A in reconstituted 43S PICs (Yu et al. 2009). This work revealed that the CTT extends into the P site, threading under the Met-tRNA_i^Met, in a configuration that would obstruct Met-tRNA_i^Met binding to the P site in the manner observed in bacterial 70S complexes. Hence, it was concluded that AUG recognition and formation of the closed complex would likely require removal of the eIF1A CTT from the P site. Indeed, we argued previously that the eIF1A CTT would be ejected from the P site on AUG recognition (Fekete et al. 2007), based on its physical displacement from eIF1 in the PIC (Maag et al. 2005) and its AUG-dependent functional interaction with eIF5 (Maag et al. 2006). It seems likely, therefore, that the SEs sterically block the P_in mode of TC binding in addition to stabilizing the P_out conformation (Fig. 7A). This idea is very attractive because SE mutations would then facilitate the P_out-to-P_in transition at UUGs (conferring Sui⁻ phenotypes) in two ways: by destabilizing P_out and also removing a steric impediment to P_in (Fig. 7B).

Combining the results of Yu et al. (2009) with our finding that SE elements promote TC binding to the scanning-conducive conformation of the PIC, we propose that, in the P_out state, the SEs occupy the P site in a manner that blocks accommodation of the initiator ASL without preventing the codon–anticodon interactions required during scanning (Fig. 7D). The SEs might interact directly with the anticodon or ASL of the initiator to stabilize TC bound in this scanning conformation of the PIC. Pairing with the AUG start codon would lead to ejection of the SEs from the P site and allow more extensive P-site engagement of the initiator ASL in the P_in conformation, as depicted in Figure 7E. The P_in conformation might resemble the “30S P/I” state visualized in a cryo-EM model of the bacterial 30S initiation complex (Simonetti et al. 2008), and further conformational changes would presumably occur on subunit joining before reaching the classical P-site binding of tRNA seen in bacterial 70S structures (Berk et al. 2006; Korostelev et al. 2006; Selmer et al. 2006). Our model in Figure 7, D and E, fits with the notion that accommodating initiator tRNA in different intermediate states is a key feature of the small ribosomal subunit (Simonetti et al. 2008).

We can readily incorporate the functions of eIF1A SI elements into our model simply by proposing that they oppose the SEs and destabilize the P_out state to which TC initially binds (Fig. 7A). This can explain our finding that SI mutations partially suppress the defective TC binding conferred by the SE₁*,SE₂* mutant in vitro and, consistently, suppress the Gcd⁻ phenotypes of SE mutations in vivo (Fig. 7C). At the same time, SI mutations would shift the equilibrium back from P_in to P_out, promoting the scanning conformation of the PIC and suppressing UUG initiation in Sui⁻ mutants; i.e., their Ssu⁻ phenotypes (Fig. 7C). The function of the wild-type SI elements in destabilizing the P_out state can be viewed as a driving force that facilitates the P_out-to-P_in transition required for AUG selection. Because the NTT also appears to interact directly with the P site (Yu et al. 2009), it might play a direct role in antagonizing initiator binding to the P_out state (Fig. 7D).

Results of our previous studies suggested that TC binds initially to the open conformation of the PIC, stabilized synergistically by eIF1 and eIF1A (Passmore et al. 2007). Hence, it is plausible that the SEs could stimulate TC loading by stabilizing the open conformation of the 40S rather than interacting directly with the initiator. Indeed, we found that inactivation of the SEs accelerates the rate of 60S subunit joining, which is thought to proceed only from the closed conformation of the 40S. Remarkably, SI substitutions had the opposite effect and partially suppressed the more rapid 60S joining conferred by the SE substitutions, as would be expected if the wild-type SIs antagonize the open conformation as the means of promoting start codon recognition. Thus, it appears likely that the SEs and SIs regulate the rate of TC loading, at least partly, by their opposing effects on the open-to-closed conformational transition of the 40S subunit. However, it is noteworthy in this connection that eIF1 overexpression did not suppress the Gcd⁻ phenotype of an SE mutation even though it suppressed the Sui⁻ phenotype (Fig. 6C,D). As eIF1 promotes the open conformation, one might expect that overexpressing eIF1 would rescue TC binding in SE mutants if this defect arises only from reduced occupancy of the open state. Hence, the SEs likely stimulate TC loading by a second, possibly direct, mechanism in addition to promoting the open conformation of the 40S subunit.

In its simplest formulation, our model posits that the SEs in eIF1A help to recruit TC in the open, scanning conformation and block initiator accommodation at non-AUGs, whereas SIs drive the transition to the closed state in which the initiator is fully engaged in the P site at AUG codons (Fig. 7). IF1, the bacterial ortholog of eIF1A, lacks the CTT, NTT, and helical domain (Supplemental Fig. S6; Carter et al. 2001), whereas the archaeal ortholog (aIF1A) lacks only the C strand and unstructured CTT (Supplemental Figs. S7, S8). Similar to bacteria, it appears that many mRNAs in archaea use the Shine-Delgarno (SD) sequence upstream of the start codon to recruit the 30S subunit directly to the initiation region (Dennis 1997; Londei 2005), obviating the scanning mechanism. Thus, in bacteria and archaea, the presence of SD-facilitated AUG selection is correlated with the absence of SEs, consistent with the role of these eukaryotic elements in promoting a scanning-competent intermediate in the initiation pathway (Fig. 7D). Considering that archaea resemble eukaryotes in employing a TC with aIF2-GTP for initiator recruitment (Pedulla et al. 2005), the helical domains in aIF1A/eIF1A might also play a role in promoting TC binding in the P_in state. Recent in vitro findings suggest that bacterial IF1 functions in stabilizing a conformation of the 30S initiation complex that is incompatible with subunit joining, which can be overcome by a favorable SD sequence (Milon et al. 2008). Thus, although IF1 lacks the SE elements, it carries out one of the functions we ascribe to the SE, of stabilizing a small subunit conformation incompatible with subunit joining. This raises the possibility that a region of the OB fold—the main structural element shared between eIF1A and IF1—could augment this aspect of SE function.

Materials and methods

Plasmids and yeast strains

Plasmids and yeast strains used in this study are listed in Supplemental Tables S1 and S2, respectively. Descriptions of plasmid constructions and site-directed mutagenesis are given in the Supplemental Material. The his4-301 yeast strain H3582 [MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301 (ACG) tif11Δ p3392 <TIF11, URA3>] was transformed with single-copy or hc LEU2 plasmids harboring various TIF11 alleles on SC-L medium, and the resident TIF11⁺ URA3 plasmid p3392 was evicted by selection on 5-FOA medium to obtain the mutant strains listed in Supplemental Table S2.

Biochemical assays with yeast extracts

Assays of β-galactosidase activity in WCEs were performed as described previously (Moehle and Hinnebusch 1991). Measurements of luminescence in WCEs were conducted essentially as described (Dyer et al. 2000). For Western analysis, WCE extracts were prepared by trichloroacetic acid extraction as described previously (Reid and Schatz 1982), and immunoblot analysis was conducted as described (Olsen et al. 2003).

Biochemical assays in the reconstituted yeast system

Reagent preparation is described in the Supplemental Material. Fluorescence anisotropy measurements to determine K_d values for 40S binding of eIF1A were performed as described previously using wild-type eIF1A Fluorescein-labeled at the C terminus and competing its 40S binding with unlabeled eIF1A mutants (Maag and Lorsch 2003; Maag et al. 2006). For all experiments, buffer conditions were 30 mM HEPES (pH 7.4), 100 mM potassium acetate (pH 7.4), 3 mM MgCl₂, and 2 mM dithiothreitol. TC binding was measured by gel mobility shift assays as described previously (Algire et al. 2002). Kinetics of ribosomal subunit joining in the presence of wild-type or mutant eIF1A was measured by light scattering on an SX.180MV-R stopped-flow fluorometer (Applied Photophysics) (Acker et al. 2009).