β-Hairpin Loop of Eukaryotic Initiation Factor 1 (eIF1) Mediates 40 S Ribosome Binding to Regulate Initiator tRNA^Met Recruitment and Accuracy of AUG Selection in Vivo ^{* ♦}

Buffers and Reagents

For all experiments, the reaction buffer was 30 mm Hepes-KOH (pH 7.4), 100 mm potassium acetate (pH 7.4), 3 mm magnesium acetate, and 2 mm dithiothreitol. The composition of the enzyme buffer was 20 mm Hepes-KOH (pH 7.4), 100 mm KOAc (pH 7.4), 2 mm DTT, and 10% glycerol.

Initiation factors eIF1A and eIF1 WT and mutant variants of this protein were purified using the IMPACT system (New England Biolabs) as described previously (19). eIF1 WT and mutant proteins were labeled at their C termini with cysteine-lysine-fluorescein dipeptide, using the expressed protein ligation system as described previously (20). His-tagged eIF2 was overexpressed in yeast and purified as described (19). 40 S subunits were purified as described previously (19). The sequences of the model mRNAs were 5′-GGAA(UC)₇UAUG(CU)₁₀C-3′ and 5′-GGAA(UC)₇UUUG(CU)₁₀C-3′. Yeast tRNA_i^Met was synthesized from a hammerhead fusion template using T7 polymerase transcription and charged with [³⁵S]Met or unlabeled methionine as described previously (19).

Fluorescence Anisotropy Experiments

Fluorescence anisotropy measurements of equilibrium binding constants (K_d) were performed using a T-format Spex Fluorolog-3 (J. Y. Horiba) as described previously (20). The excitation and emission wavelengths were 497 and 520 nm, respectively. The data were fit with a hyperbolic binding equation describing the binding of fluorescently labeled eIF1 mutants to 40 S subunits to give K_d values (20). In competition experiments with unlabeled eIF1, the data were fit with a quadratic equation describing the competitive binding of two ligands to a receptor (20).

Ternary Complex Binding Experiments

Measurement of kinetics of TC binding to 40 S subunits was monitored using a native gel shift assay as described previously (9, 19, 21). TC assembled with 0.5–1 nm [³⁵S]Met-tRNA_i was mixed with 10–40 nm 40 S subunits, eIF1 (1 μm), eIF1A (1 μm), and mRNA (10 μm). Binding of labeled TC was stopped by adding a chase of excess (≥300-fold) unlabeled TC at different times. The k_obs values were calculated by fitting the data with a single exponential equation. To determine the association rate constants (k_on), the k_obs values were measured at different 40 S concentrations; the slope of the plot of k_obs versus [40 S] yielded k_on (21).

To determine rate constants for TC dissociation from PICs, 43 S·mRNA complexes were formed using [³⁵S]Met-tRNA_i as described above. A chase of excess (≥300-fold) unlabeled TC was added, and the fraction of bound and labeled TC was monitored over time on native gels. Curves were fit to a single exponential equation, to determine the dissociation rate constant (k_off).

Genetic Evidence That Conserved Basic Residues in Helix α1 of Yeast eIF1 Mediate Critical Contacts with the 40 S Subunit in Vivo

The crystal structure of the Tetrahymena 40 S subunit bound to eIF1 identified three basic residues making direct contacts with 18 S rRNA residues, namely Arg-29 in loop-1 and Lys-52 and Lys-53 in helix α1 (12). Consistent with critical roles in eIF1 binding to the ribosome, these basic residues and surrounding amino acids are among the most highly conserved residues in eIF1 from diverse eukaryotes (Fig. 2A). Their locations in the 40 S·eIF1 complex are depicted in Fig. 2B, labeled with their amino acid positions in Saccharomyces cerevisiae eIF1, i.e. Arg-36, Lys-59, and Lys-60. We predicted that altering these basic residues in yeast eIF1 to negatively charged Asp or Glu residues would weaken 40 S binding by eIF1 and confer a Sui⁻ phenotype. Indeed, we previously reported a Sui⁻ phenotype for the K60E substitution in α1 (16). Here, we extended our analysis to include other substitutions of Lys-60 and mutations of the adjacent, invariant Lys residue at position 59.

To this end, the appropriate mutations were generated in a SUI1 allele (encoding yeast eIF1) under its native promoter on a single copy LEU2 plasmid; the resulting plasmids were introduced into a his4-301 sui1Δ strain containing WT SUI1 on a URA3 plasmid, and the latter SUI1⁺ plasmid was evicted by counter-selection on medium containing 5-FOA (25). Because the mutant his4-301 mRNA lacks an AUG start codon, the parental SUI1⁺ his4-301 strain cannot grow on medium lacking histidine (or containing only 1% of the normal histidine supplement), whereas Sui⁻ mutations restore growth on −His media by increasing initiation at the third (UUG) codon in his4-301 mRNA. Compared with the K60E substitution analyzed previously (16), we found that the K60D substitution confers an even stronger growth defect on histidine-replete (+His) medium and more pronounced growth on −His medium in the his4-301 strain; however, the K60A substitution did not alter growth on either medium (Fig. 3A). Phenotypes indistinguishable from K60E were observed for Glu and Asp substitutions at residue Lys-56 (Fig. 3A), located one helical turn upstream of Lys-59 and Lys-60 in helix α1. Although Lys-56 is not predicted to make direct contacts with the 40 S, it lies on the same face of α1 and is close to the predicted eIF1:40 S interface (Fig. 2B) (12). The K60D, K60E, K56D, and K56E substitutions all conferred an increased frequency of UUG initiation, detected by assaying matched HIS4-lacZ fusions harboring AUG or UUG start codons, demonstrating that they are bona fide Sui⁻ mutants. Consistent with the His⁺ phenotypes (Fig. 3A), the largest increase in the UUG/AUG initiation ratio (8-fold) was observed for K60D (Fig. 3B) (16). Although the K60A substitution did not confer a His⁺ phenotype (Fig. 3A), it produced a significant increase in the UUG/AUG initiation ratio, albeit of smaller degree than for the other variants. Hence, removing a positive charge at residue 60 is sufficient to reduce initiation accuracy, and introducing a negative charge evokes a similar defect of greater magnitude.

Western analysis of WCEs revealed that these sui1 mutants all contain higher than WT levels of eIF1 (normalized to the eIF2Bϵ subunit Gcd6) (Fig. 3C). The acidic substitutions of Lys-56 or Lys-60 evoked ∼4–5-fold increases, whereas the relatively weaker Sui⁻ variant K60A produced only an ∼3-fold increase in eIF1 expression. Previously, we established that the elevated expression of Sui⁻ eIF1 mutants results from diminished discrimination against the poor sequence context of the SUI1 AUG start codon with attendant increased translation of SUI1 mRNA (16). Considering that overexpression of eIF1 does not confer a Sui⁻ phenotype, and actually suppresses the Sui⁻ phenotypes of mutations in other initiation factors (2), we conclude that the Lys-56 and Lys-60 substitutions confer Sui⁻ phenotypes by altering eIF1 function rather than its expression.

Interestingly, combining the K56E and K60E mutations (K56E,K60E) was lethal, as was substitution of highly conserved Lys-59 with Glu, Asp, or Ala, as indicated by the fact that the SUI1⁺ URA3 plasmid could not be evicted from transformants harboring these sui1 alleles to permit growth on 5-FOA medium (Fig. 3D). Western analysis of His₆-tagged versions of these eIF1 proteins in viable cells that additionally harbor untagged SUI1⁺ (to maintain viability) revealed that the lethal mutations did not reduce eIF1 expression below the WT level (Fig. 3E). Thus, eliminating the positively charged side chain of Lys-59, or introducing negatively charged side chains at both Lys-56 and Lys-60 simultaneously, confers a lethal reduction in eIF1 function. These findings support the idea that Lys-59 and Lys-60 of helix α1 make critical contacts with 18 S rRNA (Fig. 2B) and suggest that Lys-59 is more important than Lys-60 for 40 S interaction with yeast eIF1, as even the Ala substitution at that position is lethal.

Genetic Evidence That the Sui⁻ and Lethal eIF1 Mutations in Helix α1 Provoke Inappropriate Dissociation of eIF1 from 40 S Subunits

We reasoned that if the lethality or Sui⁻ phenotypes of substitutions of Lys residues in α1 result from reduced eIF1 binding to the 40 S subunit and inappropriate eIF1 dissociation from the PIC at non-AUG codons, these phenotypes should be suppressed by overexpressing the mutant eIF1 proteins to restore high level 40 S occupancy by mass action (5). Consistent with this prediction, overexpressing the eIF1-K56E and eIF1-K60E variants from hc plasmids diminished the Slg⁻ phenotypes on +His medium and His⁺/Sui⁻ phenotypes on −His medium that they confer when expressed from sc plasmids (Fig. 4A, +His; −His). Similar results were observed for another Sui⁻ eIF1 substitution we obtained by random mutagenesis of SUI1, I3N, which alters a single residue in the unstructured NTT (Fig. 4A). (Fig. 4A reveals that overexpressing WT eIF1 confers a modest Slg⁻ phenotype on +His medium not seen on overexpression of the I3N, K60E, or K56E variants. Presumably, an excess of WT eIF1 interferes with some aspect of initiation in a manner eliminated by these loss-of-function substitutions.) Overexpression of the mutant eIF1 proteins also partially suppressed the elevated UUG/AUG initiation ratios of HIS4-lacZ expression conferred by the sui1 mutations (Fig. 4C). Western analysis showed that the I3N variant is expressed at WT, or slightly less than WT, levels in different transformants harboring the sc sui1-I3N plasmid (data not shown) and that all three eIF1 mutant proteins were overproduced in yeast when expressed from hc plasmids (Fig. 4B and data not shown).

Interestingly, overexpressing the sui1 alleles K59A and K56E,K60E overcame their lethality (Fig. 4, D and E), whereas the mutations that replace the crucial positive charge at Lys-59 with negatively charged Asp or Glu (K59E and K59D) remained lethal despite high level overexpression (Fig. 4, D and E). Although cells overexpressing the K59A and K56E,K60E mutants are viable, they display residual growth defects on +His medium, particularly at 37 °C, and His⁺/Sui⁻ phenotypes on −His medium (Fig. 4F); and they also elevated UUG/AUG initiation ratios (Fig. 4G). We presume that high level expression of these variants does not restore a WT level of 40 S binding. The results of the dosage suppression experiments in Fig. 4 support the idea that the eIF1 substitutions provoke inappropriate eIF1 dissociation from the 40 S subunit in a manner that elevates recognition of near-cognate UUG codons and, in the case of Lys-59 substitutions, provokes a lethal reduction in canonical AUG initiation as well.

We showed previously that Ala substitution of residues 17–21 in the NTT of eIF1A confers an Ssu⁻ phenotype in vivo, partially suppressing the Sui⁻ and Slg⁻ phenotypes of Sui⁻ mutations affecting eIF2β (SUI3–2), eIF5 (SUI5), and eIF1A itself (6, 7). Interestingly, the 17–21 substitution also reduces the rate of eIF1 dissociation from reconstituted PICs in vitro (5). Hence, we reasoned that the 17–21 mutation should suppress the His⁺/Sui⁻ and Slg⁻ phenotypes of the eIF1 mutations described above if they provoke abnormally rapid dissociation of eIF1 from PICs in vivo. To test this prediction, we generated sui1Δ strains harboring the chromosomal gene encoding eIF1A (TIF11⁺) placed under the GAL1 promoter, plasmid-borne tif11-17-21 or TIF11⁺, and mutant or WT sui1 alleles on a separate plasmid. When cultured continuously on glucose medium, where the GAL1 promoter is repressed, the plasmid-borne tif11-17-21 or TIF11⁺ allele provides the only source of eIF1A in the cell (16).

In accordance with our prediction, the tif11-17-21 mutation partially suppresses both the Slg⁻ and His⁺/Sui⁻ phenotypes conferred by the viable K60E, K56E, and I3N mutations in eIF1 (Fig. 5A). (Note that suppression of growth on −His for the K56E mutant is obscured by the improved growth on +His medium afforded by the 17–21 mutation in this double mutant.) The 17–21 mutation also diminishes the UUG/AUG initiation ratios conferred by these four Sui⁻ mutations, although the I3N substitution is inefficiently suppressed compared with the others (Fig. 5B). Strikingly, the lethality of K59A and the K56E,K60E double mutation is also suppressed by the 17–21 mutation (Fig. 5C), supporting the notion that, in otherwise WT cells, these mutations confer an intolerable reduction in eIF1 binding to the 40 S subunit, which can be overcome by stabilizing the open conformation of the PIC by an eIF1A NTT mutation. The fact that 17–21 is less effective than overexpressing the mutant eIF1 proteins in suppressing the elevated UUG/AUG ratios conferred by the viable substitutions (cf. Figs. 4C and 5B) might be explained by considering our previous finding that 17–21 exacerbates the effect of poor context at the SUI1 AUG start codon and reduces eIF1 expression (16). Reducing the abundance of the eIF1 mutant proteins would be expected to exacerbate their Sui⁻ phenotypes and thereby dampen the suppressor activity of 17–21.

Genetic Evidence That Mutations That Weaken 40 S Binding of eIF1 Decrease the Rate of TC Recruitment in Vivo

Previous biochemical findings led to the suggestion that eIF1 stabilizes the open conformation of the 40 S as a means of stimulating the rate of TC loading on the PIC (9). We reasoned that if this mechanism prevails in vivo then the Sui⁻ mutations described above, which appear to weaken 40 S binding by eIF1, should decrease the fraction of 40 S subunits in the open conformation and thereby decrease the rate at which TC re-binds to 40 S subunits engaged in reinitiation on GCN4 mRNA. This would suppress reinitiation at the inhibitory uORFs 2–4 and allow reinitiation further downstream at the GCN4 AUG codon, derepressing GCN4 translation in the absence of eIF2α phosphorylation by Gcn2 (the Gcd⁻ phenotype). Moreover, this Gcd⁻ defect should be overcome by increasing the cellular concentration of TC achieved by overexpressing all three eIF2 subunits and tRNA_i^Met from a single hc plasmid (hc TC) (11).

Supporting our predictions, the sui1 mutations I3N, K56E, and K60E all evoked derepression of a GCN4-lacZ reporter harboring all four uORFs in a manner diminished by overexpressing TC (Fig. 5D), as would be expected if this Gcd⁻ phenotype results at least partly from a reduced rate of TC binding to reinitiating 40 S subunits that can be rescued at high TC levels by mass action.

If the eIF1 mutations reduce the rate of TC loading by destabilizing the open conformation of the PIC, then their Gcd⁻ phenotypes should be suppressed by overexpressing eIF1 or by the eIF1A 17–21 substitution (7). Consistent with this prediction, overexpression of the I3N, K56E, and K60E mutants from hc plasmids efficiently suppressed the derepression of GCN4-lacZ expression that was evoked by each mutation when expressed from an sc plasmid (Fig. 5E). Moreover, the 17–21 eIF1A mutation also reduced the extent of GCN4-lacZ derepression in cells containing these same eIF1 mutants (Fig. 5F), although not to the same extent observed on their overexpression (Fig. 5E). The relatively greater suppression of the Gcd⁻ phenotypes afforded by overexpressing the eIF1 variants mirrors the findings above that mutant overexpression was more effective than 17–21 in reducing the elevated UUG/AUG ratios. Together, the genetic data support the model that mutations that weaken eIF1 binding to the 40 S subunit destabilize the open PIC conformation in vivo, reducing the rate of TC loading during reinitiation on GCN4 mRNA and also increasing the probability of inappropriate rearrangement to the closed PIC conformation at near-cognate (UUG) codons.

Genetic Evidence That Residues in eIF1 Loop-1 Are Crucial for 40 S Binding by eIF1 in Vivo

As noted above, Arg-36 corresponds to the residue in loop-1 of Tetrahymena eIF1 (Arg-29) that directly contacts 18 S rRNA in the eIF1·40 S complex (12) (Fig. 2, B and C); however, no substitutions affecting loop-1 have been described. Based on findings above, we predicted that eliminating the positively charged side chain in Arg-36 would impair cell growth and confer a Sui⁻ phenotype. Surprisingly, however, although replacing Arg-36 with acidic Asp produced a modest reduction in growth at 37 °C, the mutant grew indistinguishably from WT on −His medium (suggesting the absence of a Sui⁻ phenotype), and the R36G and R36A substitutions had no mutant phenotypes whatsoever (Fig. 6A, left panel). The same results were obtained for substitutions in residues Asn-34 and Gly-35 located immediately N-terminal of Arg-36, including N34A, N34E, G35R, and G35E (Fig. 6A, left panel). Similarly, Ala substitution of the adjacent basic residue Lys-37 produced a noticeable growth defect at 37 °C, but was His⁻ (Fig. 6A, left panel). However, the K37E replacement is lethal (Fig. 6B, left panel), even though Western analysis of His₆-tagged eIF1-K37E revealed a higher than WT level of its expression (Fig. 6C), indicating that Lys-37 carries out an essential eIF1 function.

Interestingly, although none of the substitutions of Arg-36 or nearby residues conferred His⁺ phenotypes, we found that all but the G35E and G35R variants evoke higher than WT eIF1 protein expression, with the smallest increase observed for the N34A mutant (Fig. 6D). As overexpression is a hallmark of eIF1 Sui⁻ mutants (16), we analyzed the UUG/AUG initiation ratios for these mutants. Except for G35E and G35R, we observed significant increases in UUG initiation for all the loop-1 substitutions (Fig. 6E), which were generally of lesser magnitude than those given by the Lys-56 and Lys-60 substitutions discussed above (Fig. 3B). Moreover, overexpressing the K37E protein restored cell viability (Fig. 6B, right panel), but the mutant cells displayed residual Slg⁻, temperature sensitivity, and His⁺/Sui⁻ phenotypes (Fig. 6F), plus a dramatically elevated UUG/AUG initiation ratio (Fig. 6G). These findings support the possibility that reducing the positive charge, or adding a negative charge, in loop-1 decreases 40 S binding of yeast eIF1, with Lys-37 substitutions having the greatest impact.

Although the residue corresponding to Lys-37 of yeast eIF1 is located in strand β2 rather than loop-1 of Tetrahymena eIF1 (residue Lys-30, Fig. 2C) (12), Lys-37 resides within loop-1 of yeast eIF1, which encompasses residues ³²QRNGRKT³⁸ (24) and is thus five residues larger than the two-residue loop-1 of Tetrahymena eIF1. Noting that Arg-33 is a third basic residue in loop-1 of yeast eIF1, we extended our analysis to include Ala or Glu substitutions of this residue. Interestingly, the Ala substitution confers strong Slg⁻ and His⁺/Sui⁻ phenotypes (Fig. 6A, right panel) and a markedly increased UUG/AUG initiation ratio (Fig. 6E). The R33E substitution by contrast is lethal, despite being expressed at higher than WT levels, and further overexpressing it does not restore cell viability (data not shown). Thus, considering that both R33E and K37E are lethal, whereas R36E is viable, it is possible that Arg-33 or Lys-37 carries out the critical role of Arg-29 in Tetrahymena eIF1 of contacting a specific uridine residue in helix 44 of 18 S rRNA (12). Alternatively, examination of the Tetrahymena 40 S·eIF1 structure suggests that all three of these residues could make separate contacts with adjoining rRNA residues (Fig. 2C) (12), in which case contacts made by Arg-33 and Lys-37 of yeast eIF1 could be more critical than that involving Arg-36 (see below).

Evidence that altering the loop-1 residues confers Slg⁻ and Sui⁻ phenotypes by weakening 40 S binding by eIF1 came from our finding that overexpression diminished the Slg⁻ phenotypes at 37 °C conferred by the R36E, K37A, and R33A mutations (Fig. 7A) and reduced the elevated UUG/AUG initiation ratios conferred by all of the loop-1 substitutions that confer increased UUG initiation (Fig. 7C). We verified that all of the loop-1 mutant proteins were overexpressed from the hc plasmids (data not shown). Similar results were obtained when the eIF1 mutations were combined with the 17–21 mutation (Fig. 7, B and D), except that suppression by 17–21 was relatively less complete, as observed above for other Sui⁻ mutations. As would be expected for mutations that reduce eIF1 binding to the 40 S subunit, all of the loop-1 mutations also confer Gcd⁻ phenotypes, derepressing GCN4-lacZ expression to an extent that generally correlates with the degree to which they increase the UUG/AUG ratio (cf. Fig. 7, C and E, sc SUI1 lanes), and this phenotype is likewise suppressed by overexpressing the mutant proteins and by the 17–21 mutation (Fig. 7, E and F). These findings provide strong evidence that the basic residues of β-hairpin loop-1 stabilize eIF1 binding to the 40 S subunit in vivo and thereby promote TC recruitment and accurate start codon selection.

Basic Residues in Helix α1 and Loop-1 Are Crucial for Stable eIF1 Binding to 40 S Subunits in Vitro

To provide biochemical evidence supporting the conclusion that basic residues in α1 and loop-1 mediate direct binding of yeast eIF1 to 40 S subunits, we measured the 40 S binding affinity of purified, recombinant forms of the eIF1 variants.

We first assessed whether the mutations perturbed the folding of eIF1 by comparing the dispersion of the chemical shifts of the fingerprint HSQC spectra between WT eIF1 and the mutants I3N, R33A, K37E, and K60E. The spectra for WT and the four eIF1 mutants exhibit very similar ¹H and ¹⁵N chemical shift patterns, confirming that the tertiary folds of these proteins are nearly identical (Fig. 8, A–E). These results indicate that none of the substitutions produces a notable alteration of the tertiary structure of eIF1.

The eIF1 variants were fluorescently labeled on their C termini with fluorescein, and the fraction of eIF1_FL bound to 40 S subunits alone or in the presence of a saturating concentration of eIF1A was measured at different concentrations of 40 S subunits by monitoring the change in fluorescence anisotropy. In agreement with previous results, the affinity of WT eIF1 for 40 S subunits was increased by the presence of eIF1A, lowering the K_d value into the low nanomolar range, because of thermodynamic coupling of their association with ribosomes (Fig. 9A) (20). The I3N substitution in the NTT did not significantly alter eIF1 binding to 40 S·eIF1A complexes (Fig. 9A), whereas the R33A mutation in loop-1 increased the K_d value of eIF1 for 40 S subunits in the presence of eIF1A by ∼8-fold (Fig. 9, B and C). The K60E and K37E substitutions in α1 and loop-1 more dramatically reduced the affinity of eIF1 for the 40 S·eIF1A complex, increasing the K_d values in the presence of eIF1A to >500 nm and >1 μm, respectively (Fig. 9, B and C).

To confirm these findings with unlabeled variants, we conducted competition binding assays wherein 40 S·eIF1A·eIF1_FL complexes formed with WT eIF1_FL were challenged with increasing concentrations of unlabeled WT or mutant eIF1, and the fraction of eIF1_FL bound to 40 S·eIF1A complexes was measured at each concentration of competitor. In accordance with the direct binding assays, even at very high concentrations, the K60E and K37E variants were unable to compete with WT eIF1_FL for binding to 40 S·eIF1A complexes (Fig. 9D). Finally, we conducted competition binding assays using WT eIF1_FL bound to 43 S·mRNA (UUG) complexes, formed with 40 S subunits, eIF1_FL, eIF1A, TC, and an unstructured model mRNA with a UUG start codon. In lacking an AUG codon, this 43 S·mRNA (UUG) complex is considered to be a model of the scanning PIC. The results showed that the K60E and K37E mutants could not compete effectively with WT eIF1_FL for binding to 43 S·mRNA (UUG) PICs (Fig. 9E). We conclude that conserved basic residues in helix α1 and loop-1 are crucial for stable binding of eIF1 to a reconstituted complex representing the scanning PIC.

I3N Substitution Impairs Effector Functions of the eIF1 NTT in TC Recruitment and Start Codon Selection

Having found that the I3N substitution does not affect the affinity of eIF1 for the 40 S subunit, we hypothesized that it impaired an important function of the NTT in regulating start codon recognition. We first investigated the effect of I3N on the rate of TC binding to PICs reconstituted with eIF1A, 40 S subunits, a model mRNA with AUG or UUG start codon, and saturating amounts of either WT or mutant eIF1. The TC was assembled with [³⁵S]Met-tRNA_i^Met, and its binding to the 40 S subunit was monitored by electrophoretic mobility shift (21). As shown in Fig. 10A, the I3N substitution produced an obvious reduction in the observed rate of TC binding to complexes formed with the AUG mRNA, and similar results were observed for the UUG mRNA (data not shown). By measuring the k_obs at different 40 S concentrations, we determined that I3N decreased the k_on by a factor of 2 for mRNAs containing either AUG or UUG start codons (Fig. 10, B and C). This reduced rate of TC binding is consistent with the Gcd⁻ phenotype of the I3N substitution in vivo.

We also examined the stability of TC binding to the PIC by measuring the rate of dissociation of ³⁵S-labeled TC from pre-formed 43 S·mRNA complexes using an excess of unlabeled TC as a “chase.” Consistent with previous results (21), the rate of TC dissociation is substantially lower with an AUG versus UUG start codon in the mRNA, with k_off values of 0.07 and 0.40 h⁻¹ at AUG and UUG, respectively (Fig. 10, D and E). This difference is thought to reflect the greater stability of the P_IN state of TC binding with an AUG versus the near-cognate UUG start codon. Interestingly, I3N considerably decreases k_off for both start codons (Fig. 10, D and E). The increased stability of the P_IN state at the UUG codon inferred from these data is likely instrumental in the increased UUG initiation rate conferred by the I3N substitution in vivo. Perhaps the increased stability of the AUG complex for the I3N mutant would not similarly increase the initiation rate at AUG codons because the P_IN state at AUG for WT eIF1 is already optimal for efficient initiation.