Large ribosomal subunit, eIF5B, Met-tRNAiMet and mRNA cooperate to complete accurate initiation

Recognition of a start codon by the first aminoacyl-tRNA (Met-tRNAiMet) determines the reading frame of messenger RNA (mRNA) translation by the ribosome. In eukaryotes, the GTPase eIF5B collaborates in the correct positioning of Met-tRNAiMet on the ribosome in the later stages of translation initiation, gating entrance into elongation. Leveraging the long residence time of eIF5B on the ribosome recently identified by single-molecule fluorescence measurements, we determined the cryoEM structure of the naturally long-lived ribosome complex with eIF5B and Met-tRNAiMet immediately before transition into elongation. The structure uncovered an unexpected, eukaryotic specific and dynamic fidelity checkpoint implemented by eIF5B in concert with components of the large ribosomal subunit. One sentence summary CryoEM structure of a naturally long-lived translation initiation intermediate with Met-tRNAiMet and eIF5B post GTP hydrolysis.

reconstructions have described the architectures of the 80S IC bound with eIF5B:GDPCP at medium 48 resolution, representing snapshots of the assembly prior to eIF5B GTP hydrolysis (12,13). In addition, 49 Met-tRNAi Met has sequence features that allow it to bind directly to the P site of the ribosome during 50 initiation, unlike elongator tRNAs (14)(15)(16). Here, guided by the kinetics determined by our previous single-51 molecule fluorescence measurements, we performed cryoEM analysis of the on-pathway initiation 52 complexes to provide high-resolution information on the molecular mechanism by which eIF5B escorts the 53 initiator tRNA into the ribosomal P site and gates the transition from initiation to elongation. 54 Leveraging the slow dissociation of eIF5B from the 80S IC in the native initiation pathway (average 55 lifetime of the eIF5B-bound 80S state is 30-60 s at 20ºC) (11), we prepared and froze samples at a pre-56 steady-state reaction timepoint (40 s, Fig. 1B, Fig. S1 and Methods) corresponding to when ~60% of the 57 assembled 80S should be bound with eIF5B (80S IC) after mixing 48S PICs with eIF5B:GTP and 60S. 58 Image processing and 3D classification in Relion3 (17,18)  is not limited to the 73ACCA76-Met of tRNAi Met , as the whole acceptor stem is distorted compared 72 with its position in an elongation tRNA ( Fig. 2B and C). The cluster of G-C base pairs of the 73 acceptor stem specific to Met-tRNAi Met seems to play a pivotal role in this context, allowing a 74 specific distortion of this stem as the Met-tRNAi Met simultaneously interacts with eIF5B and the 75 start-codon ( Fig. 2A and B and Fig. S5). Basic residues of the domain IV of eIF5B make specific interactions with this G-C base pairs cluster (21); Arg955 of eIF5B contacts the base of G70 of 77 Met-tRNAi Met from the major groove of the acceptor stem ( Fig. 2E and F). Although mutating 78 Arg955 to Ala in eIF5B did not substantially alter the growth rate of yeast in rich medium, the 79 mutant strain could not grow under amino acid starvation conditions (Fig. 2G), suggesting 80 impaired GCN4 expression that could result from ribosomes scanning past the start codon (termed 81 "leaky scanning") of the stimulatory first upstream open reading frame (uORF1) in the mRNA.

82
Thus, the interaction between eIF5B Arg955 and Met-tRNAi Met G70 may also play an important 83 role in start-site selection. Consistently, deletion of eIF5B DIV has also been shown to enhance 84 leaky scanning in yeast (19).

85
In contrast, the anticodon arm of Met-tRNAi Met features a configuration very similar to that 86 described for a peptidyl-tRNA in an elongation, canonical state (Fig. 2D, Fig. S5B and Movie S1)  Thus, both G-C base pair clusters at the acceptor stem and the anticodon arm of Met-tRNAi Met are 94 essential to allow a specific conformation of the Met-tRNAi Met on the P site in both the early (21) 95 and late stages of initiation (14). 96 For the mRNA, we could unambiguously identify six nucleotides including the AUG start 97 codon and the three nucleotides immediately upstream (Fig. 3). Only weak densities ascribable to 98 the A site codon could be observed, and no ordered RNA density could be identified at the entry 99 nor exit sites of the mRNA channel on the 40S. This is in marked contrast with 48S PIC structures 100 from earlier initiation stages, in which long stretches of the mRNA from the entry to the exit sites 101 were well resolved (5), pinpointing a less prominent role of mRNA/40S interactions once initiation 102 has progressed into its later stages.

103
The codon-anticodon interaction observed in our reconstruction is virtually identical to a  The high quality of the map around the PTC region allowed a precise modeling, revealing 117 specific contacts of residues of eIF5B DIV with the four terminal bases of the Met-tRNAi Met as 118 well as with ribose and phosphate backbone atoms (Fig. 4). Specifically, the base of Met-tRNAi Met 119 nucleotide A76 is narrowly monitored by eIF5B residues Glu921 and His924, which anchors the 120 adenine moiety to eIF5B DIV ( Fig. 4B and C). In this orientation, the methionyl group esterified to the 3'OH of the A76 ribose is directed towards a hydrophobic "pocket" formed by the surface 122 of eIF5B around residue Ile874 ( Fig. 4C and D and Fig. S6).

123
Intriguingly, this "hydrophobic pocket" is capped by a loop of ribosomal protein uL16, a 124 component of the 60S (Fig. 4B-D and Fig. S6), which is conserved in yeast and humans. This uL16 125 loop, formed by residues 100 to 120, is disordered in reported elongation 80S complexes, but is 126 well resolved in our reconstruction, allowing its modelling and refinement. Notably, this loop is 127 stringently checked at the last step of 60S biogenesis by a sophisticated cellular machinery that 128 blocks 60S exporting to the cytoplasm if its integrity is compromised (23). No clear function has 129 been assigned for this loop that would justify such a conserved and costly cellular machinery (24).

130
Here, residues Ser104 to Arg110 of this loop are located in close proximity to the either eIF5B or uL16 residues could be observed, which points towards an overall chemical 136 requirement for an amino acid attached to an initiator tRNA in terms of size and hydrophobicity 137 rather than a specific amino acid identity. Indeed, when tRNAi Met is mis-acylated with glycine, 138 Gly-tRNAi Met is active in 48S PIC and 80S formation (Fig. S7). However, subunit joining rate is 139 ~9-fold slower and the subsequent transition to elongation became ~3-fold faster. Both effects 140 might reflect weakened interactions between tRNAi Met and eIF5B/uL16 when the methionine 141 moiety was replaced by the smaller, non-hydrophobic glycine residue.

142
Analysis of the G domain of eIF5B that binds GTP reveals clear density for a bound 143 nucleotide that however lacks features compatible with the presence of a γ-phosphate, implying that the nucleotide state is either GDP or GDP-Pi (25, 26). Moreover, switch I of eIF5B is 145 disordered, which differs from the ordered state prior to GTP hydrolysis (25, 27). Thus, our 146 reconstruction represents an intermediate of the 80S IC right before its transition to the elongation-147 competent state, but post GTP hydrolysis. This is further supported by the fact that the ribosomal 148 inter-subunit configuration observed here is different from the 80S IC state prior to GTP hydrolysis 149 (Fig. S8). A ~3º counterclockwise rotation of the 40S in relation to the 60S was observed in the 150 pre-GTP hydrolysis state, which was coupled with apparent 40S head swivel ( Fig. S8 (12, 13, 28)). 151 However, the eIF5B-bound 80S IC observed here presents a configuration very similar to a 152 canonical non-rotated 80S complex, with virtually no rotation of the small subunit and minimal 153 swiveling of the 40S head (Fig. S8). Thus, the 80S IC will reconfigure its conformation after GTP 154 hydrolysis to a state that is more similar to the non-rotated elongation-competent state, and this 155 reconfiguration is coupled to the structural rearrangements of eIF5B.

156
In summary, our structural analysis of the pre-steady-state initiation complexes identified
80S:eIF5B complex assembly on model mRNA Previously by applying single-molecule fluorescence microscopy methods to a purified, reconstituted yeast translation system, we have revealed that eIF5B is the gating factor during the transition from eukaryotic translation initiation to elongation (1). We preformed post-scanning 48S preinitiation complexes (48S PICs, wherein 40S was labeled with a Cy3 dye) which were immobilized on zero-mode waveguides (ZMWs) imaging surface via a biotin at the 3'end of the mRNAs (Fig. S1, A) (1). After washing away unbound components, 60S (Cy5-labeled), eIF5B (Cy5.5-labeled) and the first elongator Phe-tRNA Phe (Cy3.5-labeled) in ternary complex with the elongation factor eEF1A and GTP (Phe-TC) were delivered along with required eIFs to start the reaction. By directly illuminating all the fluorescent dyes, we could monitor, in real time, the order of molecular events occurring during the late translation initiation stages and its transition to elongation (Fig. S1, B) (1). This allowed us to measure and determine the occupancy times of eIF5B on newly formed 80S complexes on the native reaction pathway. In particular, for the model mRNA-Kozak, the mean time of 60S joining to form an 80S was ~16 s with the eIF5B occupancy time on the 80S ~34 s at 20ºC and 3 mM free Mg 2+ ; whereas these values were ~3.6 s and 68.7 s, respectively, at 20ºC and 10 mM free Mg 2+ (1). Simulating the kinetic curves from these two reactions provided the information about the time-evolution of the populations of different complexes (Fig. S1, C) (1). To aid the reconstruction of a high resolution structure of the onpathway eIF5B-bound 80S prior to its transition to elongation without the need of nonhydrolysable GTP analogs or mutants, we decided to prepare and freeze our sample at timepoint 40 s at 20ºC and 10 mM free Mg 2+ , where we would expect the eIF5B-bound 80S population accounts ~60% of the total 80S particles (Fig. S1, C).
Samples were prepared in the buffer containing 30 mM HEPES-KOH pH 7.5, 100 mM KOAc, 10 mM Mg(OAc)2 and 1 mM GTP:Mg 2+ . First, a ternary complex mixture was prepared by preincubating 3.8 μM eIF2 at 30ºC for 10 min, followed by another 5 min incubation at 30ºC after addition of 2.8 μM Met-tRNAi. Next, this ternary complex mixture was diluted to one third of the concentration, and incubated at 30ºC together with 1 μM eIF1, 1 μM eIF1A, 0.5 μM model mRNA and 0.3 μM 40S for 15 min, resulting a 48S PIC mixture. Separately, a 60S mixture was prepared by mixing 1 μM eIF5, 1μM eIF5B and 0.15 μM 60S. The 48S PIC and 60S mixtures were kept on ice before sample freezing. In parallel, the 200-mesh Quantifoil R2/1 grids (Electron Microscopy Sciences, Q250AR1) were glow-discharged for 25 s in a PELCO easiGlow glow discharger (Ted Pella, Inc.). After prewarming the samples to room temperature, 3.5 μL of the 48S PIC mixture was mixed with 3.5 μL of the 60S mixture. A 3 μL sample from the resulted mixture was applied to each grid at 21 °C and 95% humidity. The sample was vitrified by plunging into liquid ethane after 2.5 s blotting using a Leica EM GP (Leica Microsystems) plunger. Total time from combining the 48S PIC and 60S mixtures to grid freezing was ~40 s.
Single-molecule experiments comparing Met-tRNAi Met and Gly-tRNAi Met . Native yeast tRNAi Met was mis-acylated by the flexizyme dFx with Gly-DBE as described (2,3). Real-time single-molecule experiments on the ZMW-based PacBio RSII instrumentation and data analyses were performed with exactly the same methodology as previously described (1). The mean times of 60S arrival to the immobilized 48S PICs (60S arrival time) and the subsequent transition to elongation (Dt, Fig.S1, B) were determined in experiments performed with the model mRNA-Kozak and unlabeled eIF5B at 3 mM free Mg 2+ and 20ºC.
Image processing and structure determination. Contrast transfer function parameters were estimated using CTFIND4 (5) and particle picking was performed using Relion3.1(6) without the use of templates and with a diameter value of 260 Ångstrongs. All 2D and 3D classifications and refinements were performed using RELION (6)(7)(8). An initial 2D classification with a 4 times binned dataset identified all ribosome particles. A consensus reconstruction with all 80S particles was computed using the AutoRefine tool of RELION. Next, 3D classification without alignment and a mask including the inter-subunit space and the 40S head (four classes, T parameter 4) identified a class with unambiguous density for eIF5B and a tRNA in the P site (7). This class was independently processed with unbinned data, yielding high resolution maps with density features in agreement with the reported resolution. Local resolution was computed with RESMAP (9).
Model building and refinement. Models from yeast 40S, 60S (10), tRNAi and eIF5B were docked into the maps using CHIMERA (11), and COOT (12) was used to manually adjust these initial models. An initial round of refinement was performed in Phenix using real-space refinement (13) with secondary structure restraints and a final step of reciprocal-space refinement with REFMAC (14). The fit of the model to the map over-fitting tests were performed following standard protocols in the field (15).       The cumulative probability distributions of the dwell times for 60S joining (A) and the transition to elongation (Δt, B) from experiments performed at 3 mM free Mg 2+ and 20 °C with the model mRNA-Kozak and dark eIF5B. Data points were fitted to a single-exponential equation. When Met-tRNAi Met was used, the mean 60S arrival time was 14.8 s ( ± 0.4 s, 95% confidence interval), with Δt of 33.7 s (± 0.9 s, 95% confidence interval) (number of molecules analyzed = 127), similar to those values determined previously under the same experimental conditions (1). For Gly-tRNAi Met (red) the mean 60S arrival time increased to 142.9 s ( ± 6.2 s, 95% confidence interval), and Δt decreased to 11.7 s (± 0.5 s, 95% confidence interval) (number of molecules analyzed = 107). In the presence of a non-hydrolysable GTP analog, binding of eIF5B to the 80S induces a moderate degree of 40S rotation compared with a canonical state (10,18). (B) In the present reconstruction, with eIF5B in the 80S complex and hydrolyzed GTP, the 40S subunit features an almost canonical configuration with very little rotation and/or head swivel.
Movie S1. Met-tRNAi Met transition from late initiation to elongation. The interaction DIV of eIF5B establishes with the acceptor stem of Met-tRNAi Met keep the 73ACCA76-Met of the tRNA away from the PTC (blue). Upon eIF5B departure, the 73ACCA76-Met is free to accommodate in the PTC so the Met-tRNAi Met adopts a full elongation-competent conformation.