Involvement of protein IF2 N domain in ribosomal subunit joining revealed from architecture and function of the full-length initiation factor

Significance

This work reports unique insights into IF2 function during eubacterial translation initiation by addressing the function of the N domain within the structure of the full-length factor in isolated form or ribosome bound, using crystallography, SAXS, cryo-EM, fast kinetics, and single molecule fluorescence.

Abstract

Translation initiation factor 2 (IF2) promotes 30S initiation complex (IC) formation and 50S subunit joining, which produces the 70S IC. The architecture of full-length IF2, determined by small angle X-ray diffraction and cryo electron microscopy, reveals a more extended conformation of IF2 in solution and on the ribosome than in the crystal. The N-terminal domain is only partially visible in the 30S IC, but in the 70S IC, it stabilizes interactions between IF2 and the L7/L12 stalk of the 50S, and on its deletion, proper N-formyl-methionyl(fMet)-tRNA^fMet positioning and efficient transpeptidation are affected. Accordingly, fast kinetics and single-molecule fluorescence data indicate that the N terminus promotes 70S IC formation by stabilizing the productive sampling of the 50S subunit during 30S IC joining. Together, our data highlight the dynamics of IF2-dependent ribosomal subunit joining and the role played by the N terminus of IF2 in this process.

In bacteria, three translation initiation factors (IF1, IF2, and IF3) participate in the first phase of protein synthesis during which the 30S subunit binds the initiator tRNA to form codon–anticodon interactions with the mRNA. The 30S initiation complex (IC) thus formed is converted into a 70S IC by joining of the 50S ribosomal subunit (1, 2). The IFs enhance the rate of ICs formation and ensure translation accuracy. Within this process, IF2 plays a key role by promoting the recruitment and stabilization of the N-formyl-methionyl(fMet)-tRNA^fMet on the 30S IC in the peptidyl/initiator (P/I) site of the small ribosomal subunit with the help of initiation factors IF1 and IF3 (3–6). IF2 mediates subunit joining on which its GTPase activity is stimulated, leading to the formation of a 70S IC in a conformation that is productive in the first transpeptidation reaction (7). Snapshots of this process, visualized by cryo electron microscopy (cryo-EM) (3, 8–10), have localized IF2 on the ribosome and provided structural insights into 30S and 70S IC’s.

Bacterial IF2 is a multidomain protein (Figs. S1 and S2) for which the overall architecture is unknown. Unlike 70S-interacting elongation GTPases, IF2 contains a specific N-terminal (N) domain that is variable in sequence (in Escherichia coli, two copies of it comprise the bona fide N domain and the G1 domain). Furthermore, it contains a highly conserved C-terminal region with a central core comprising domains G [GTP/GDP (G) binding domain, G2] and G3 (II), and a C-terminal part consisting of domain C1 and of the initiator tRNA-binding domain C2. Recently, the crystal structure of the core part (1–363) of Thermus thermophilus IF2 was determined in different functional states (apo-protein, GTP, and GDP forms) (11). These structures gave important insights into the mechanism of nucleotide binding and GTP hydrolysis and also reveal significant structural differences with a related archaeal and eukaryotic protein a/eIF5B (11). However, because the entire C-terminal region was missing, the position of the C1 and C2 domains relative to the IF2 core domains could not be determined. The crystal structure of IF2 including the C1 domain was recently determined (12). However, the relative orientation of the C1 and C2 domains still remained unclear because the C1 domain is seen in different conformations (there are two major conformations, apo and GDP bound, in the four molecules of the asymmetric unit), and the C2 domain is not visible, probably due to structural flexibility. The cryo-EM reconstruction of the 30S IC with IF2 bound (3) suggested that C1 and C2 domains extend to reach the initiator tRNA bound in an intermediary P/I state in which the tRNA elbow is tilted toward the exit site and the CCA end is lifted and bent toward IF2. This structure highlighted the binding cooperativity of IF2 and the fMet-tRNA^fMet, which allows the precise positioning of these ligands within the 30S IC (3) and enables the productive joining of the 50S subunit to form a 70S IC after GTP hydrolysis. During this process, IF2 undergoes conformational changes that lead to release of interactions with the initiator tRNA and IF2 relocation to a position from which IF2 can leave the ribosome (2, 4, 8). The absence of a full-length structure of IF2, which would be helpful to interpret the available cryo-EM structures in more detail, has hindered a complete understanding of bacterial translation initiation at the molecular level.

The crystal structure of residues 1–363 of T. thermophilus IF2 (11) has recently revealed the fold of the N-terminal region (Fig. S2E) corresponding to both N and G1 domains of E. coli and Bacillus stearothermophilus IF2. In T. thermophilus IF2, the N-terminal portion consists of a long α-helix (H3N) covered by two small α-helices at the tip (H1N and H2N) forming a shape similar to a clenched fist; the same topology is seen in the IF2 3–467 crystal structure (12). Although its function(s) remain(s) to be clarified, it is noteworthy that this N-terminal region is present in all eubacterial IF2s but is absent or not conserved in eukaryotic and archaeal IF2-related proteins such as aIF5B. However, the length and sequence of the N domain and, at least to some extent, of G1 vary extensively between eubacterial species, with T. thermophilus IF2 harboring one of the shortest. Based on a multiple sequence alignment (Figs. S1 and S2) the N-terminal region of T. thermophilus (1–69) was found to represent a conserved core in all IF2s, usually in one or two copies. This region has a sequence similar to that of the G1 domain of E. coli IF2, right before the G domain starts; this sequence is repeated at the very N-terminal end, and the two repeats are separated by a linker that is not conserved in the IF2 family. Interestingly, the NMR structure of the N-terminal E. coli IF2 repeat (13) and the crystal structure of the 1–69 region of T. thermophilus IF2 (11) reveal an identical protein fold (Fig. S2 D and E). The number of repeats is variable in different IF2 isoforms; the long N domain of E. coli IF2α (390 amino acids) contains two copies at either end, whereas only one copy is found at the very N-terminal end of IF2β (225 amino acids). These observations raise the question of the role of the N-terminal region (i.e., of the N and G1 domains) of IF2 and of its location in the overall structure of IF2 in the isolated or ribosome-bound form.

The architecture of full-length IF2, the localization of the individual domains, and the functional role of the IF2-specific N-terminal domain are described here. The structure of full-length IF2 in solution adopts a conformation different from that observed in the incomplete crystal structures. An IF2 mutant in which the entire N-terminal region is deleted (IF2-Δ69) was designed, based on the precisely defined boundaries of this region seen in the crystal structure of the IF2 core (11), and used to elucidate the functional role of the IF2 N-terminal region. All experiments were performed in a homologous system (T. thermophilus IF2 and ribosome) with the exception of the single-molecule fluorescence experiments because of the available markers for labeling the ribosomal subunits from E. coli (14). Full-length IF2 and IF2-Δ69 were purified to homogeneity and analyzed using an integrative approach comprising small angle X-ray scattering (SAXS), cryo-EM, fast kinetics, and single-molecule fluorescence studies. The results reveal that the N-terminal part of IF2 is involved in the joining of the ribosomal subunits leading to the formation of a stable 70S IC.

Results

Overall Architecture of Full-Length IF2 Derived by SAXS Analysis.

Full-length T. thermophilus IF2 (1–571) was purified as described (11) and analyzed by SAXS (15) to obtain its overall topology in solution. In light of the tendency of IF2 to proteolyze into two fragments (1–363 and 364–571; Experimental Procedures) and to optimize sample homogeneity, which is critical for SAXS experiments, the sample was subjected to size exclusion chromatography (SEC) connected to a SAXS cell (16). The scattering pattern of full-length IF2 was measured up to a momentum transfer of Q_max = 0.50 Å^-1 (Fig. 1), and structural parameters derived, such as the radius of gyration (Rg = 37.2 ± 0.5 Å) and the maximum particle dimension (D_max = 138 ± 10 Å), suggest that the protein has a rather elongated shape (Fig. S3). The molecular envelope calculated ab initio using DAMMIF (17) shows an elongated shape with distinct regions consisting of a central bulky domain flanked by two protrusions (Fig. 1B). To address the architecture of IF2 in more detail, rigid-body refinement against the scattering data was performed with DADIMODO (18) using the crystal structures of IF2 [residues 1–363 (11) or 3–467 (12)] and the NMR structure of the C2 domain (19), which is not visible in the IF2 (3–467) crystal structure. The computed scattering patterns of all of the SAXS-compatible rigid-body refined solution models are very similar with each other and in excellent agreement with the experimental data (χ² = 1.4; Experimental Procedures). In addition, the rigid-body refined models of full-length IF2 compare well with the ab initio envelope (Fig. 1C). The model allows annotating the N- and C-terminal domains because of their different sizes, thus revealing the topological features of IF2, where these domains protrude from opposite sides of a globular core formed by the G2 and G3 domains (this is also confirmed by the SAXS analysis of IF2-Δ69, providing the same topology of the protein core; Fig. 1 D–F). In comparison with the crystal structure of apo and GDP-bound IF2 (3–467) that comprises the C1 domain (12), native IF2 appears less compact in solution because the C1 domain expands from domain G3 and is further extended by the C2 domain. In contrast, in the crystal structure in the apo or GDP conformation the C1 domain is either fanned out or in direct contact with the G domain in the area of helices 4 and 6 (Fig. 2B, Lower). These findings indicate that the C1 domain is highly flexible so that it may require additional support to be stabilized. In contrast to C1, the N domain is less flexible, but yet with an orientation differing by ∼40° between the two crystal structures (11, 12) consistent with some flexibility suggested by an NMR study (13). The more extended conformation of the C1 and C2 domains in solution can be explained by the need by the latter to reach the acceptor end of fMet-tRNA^fMet when IF2 is 30S bound (3). On the other hand, because it is not easy to imagine which role could be attributed to the N domain oriented outward of the core in the structure of isolated IF2, the structure of IF2 was also analyzed within the functional ribosome context.

Fig. 1.

Full-length model of IF2 derived from experimental SAXS data. SAXS analysis of full IF2 (A–C) and IF2-Δ69 (D–F). (A and D) X-ray scattering pattern (blue dots) and scattering curve computed from the DADIMODO refined models (red curve). (Inset) Guinier plot. (B and E) Ab initio envelopes calculated with DAMMIF. (C and E) Representative SAXS models obtained after DADIMODO rigid-body refinement superimposed to the ab initio envelope. (F) Comparison of the SAXS-models derived for native IF2 and IF2-Δ69.

Fig. 2.

Cryo-EM structure of IF2 bound in 30S and 70S ICs. (A) Structure of the 30S IC complex with the fitted model; IF1, IF2, and fMet-tRNA^fMet are indicated. (B) Segmented region of the 30S IC showing the relative position of the IF2 domains, the mRNA path, and the anti–Shine-Dalgarno region. (Lower) Illustration of the conformational switch of domain C1 between the crystal structure of the isolated apo-IF2 (3-467) (12) and the 30S-bound factor. (C) Overall architecture of IF2 as derived from the cryo-EM structure of the 30S IC and the available crystal structures of IF2 [1–363 (11); 3–467 (12)] and the NMR structure of the C2 domain (19), which is not visible in the IF2 (3–467) crystal structure, providing an molecular mechanics (MM)-refined full-length IF2 model. (Inset) Viewing angle of the 30S IC. (D) Comparison of the 30S ICs with native IF2 and IF2-Δ69. (E and F) Comparison of the 30S IC with wild type (WT) IF2 and 70S IC with IF2-Δ69.

Cryo-EM Structures of 30S and 70S ICs with IF2 and IF2-Δ69.

To address the structure of ribosome-bound IF2, four different complexes were analyzed (Fig. 2). These complexes were as follows: (i) a 30S IC containing IF1, IF2, and fMet-tRNA^fMet; (ii) an identical 30S IC in which native IF2 was replaced by the IF2-Δ69 mutant; (iii) a 70S IC that forms on joining the aforementioned 30S IC with the 50S subunit in the presence of guanosine 5'-[β,γ-imido]triphosphate (GDPNP) to prevent both GTP hydrolysis and IF2 dissociation; and (iv) an identical 70S IC containing the IF2-Δ69 mutant instead of native IF2. To increase sample homogeneity, the two 70S complexes were formed starting with 70S ribosomes instead of isolated subunits. A sample homogeneity test using cryo-EM particle sorting [based on 3D resampling and classification (3D-SC; ref. 3); Experimental Procedures] revealed that both full-length and IF2-Δ69 were present in the 30S ICs. Whereas the complexes bearing the N-terminally deleted IF2 are analyzed here, the 30S IC with full-length IF2 is the same as previously described (3), but refined using data collected from a high-resolution electron microscope (Polara) run at lower acceleration voltage for the 30S ICs to improve the image contrast and facilitate particle sorting and structure refinement of the intrinsically heterogeneous complex; e.g., the map of the 30S IC shows IF2 domains better resolved than previously (3) with C1 closer to the G domain rather than to the C2 domain. The cryo-EM structure of the 30S IC, in combination with the crystal structures of isolated IF2, now allows a full, detailed molecular interpretation of the maps (Fig. 2), whereas before only overall protein domains could be annotated (3) using the crystal structure of archaeal a/eIF5B (20). The fitting reveals that there is a large conformational change with respect to the G2/G3 core domains (Fig. 2B) involving the C1 domain that is rotated around helix H8 (serves as linker to domain G3). The C1 domain is flipped by 180° compared with the IF2 crystal structure (3–467, apo form) and is positioned closer to ribosomal protein S12 and IF1 on the 30S and is further extended by the C2 domain (Fig. 2). The rotation results in the C1 domain being folded backward along helix 8 providing a direct contact to the G domain at the level of the nucleotide binding pocket and switch II, rather than with helices 4 and 6 on the other side of the G domain as in the IF2 crystal structure (3–467). The C-terminal end of C1 is oriented toward the C2 domain [modeled using its isolated structure (19)] for connectivity reasons. Thus, whereas the C1 domain in apo or GDP conformation is either pointing away from or is contacting helices 4 and 6 of the G domain in the crystalline state of the isolated IF2 (3–467), on the 30S, the C1 domain contacts the G domain in the functionally key region of the nucleotide and switch II [visible in the 1–363 crystal structure (11)]. This contact may also explain why the switch I region is much shorter in IF2 than in other ribosomal GTPases (11) because it would otherwise sterically block this interaction. The cryo-EM based modeling of 30S-bound IF2 (Fig. 2C) is consistent with the SAXS-derived model (Fig. 1C), suggesting that the conformation of the factor is similar in the free and ribosome-bound form (Fig. S4), but more compact in the crystal structure (3–467) of the isolated factor.

Interestingly, helix H8, which is found at the beginning of domain C1, is bent in the crystal structure. This structural kink is probably related to the presence of a proline residue (Pro355, unique in T. thermophilus) that disrupts the continuity of the helix. The cryo-EM structure reveals that in the 30S IC this position is located in the vicinity of ribosomal protein S12, suggesting that S12 may possibly favor either directly or allosterically the rotated and extended conformation of domain C1 when IF2 is 30S bound. Furthermore, it is clear from the cryo-EM structure that domains C1 and C2 form a continuous, elongated bipartite structure without a long helix connecting them, in contrast to what was observed in the isolated a/eIF5B (20). Consistently, the C1 and C2 domains are rather flexible elements in the isolated factor, as illustrated by their variable conformations (12, 21). In conclusion, S12 could be the stabilizing element sought for, perhaps through a small point contact at the end of helix 8 (Fig. 2E), which allows positioning the C1 domain such that the C2 domain can stabilize the initiator tRNA on the 30S IC.

The cryo-EM structure of the 30S IC reveals a small density protruding out of the G domain that corresponds to the position of the N domain (this density is absent in the 30S IC with IF2-Δ69; Fig. 2D). The approximate orientation of the N domain on the 30S subunit can be inferred from the long α-helix (H3N) present at the beginning of this domain, whereas the rest of the domain is invisible in the map (Fig. 2 A and D). This observation could indicate some structural disorder consistent with the N domain being solvent-exposed and flexible and held in place by a single helix without additional, stabilizing interdomain contacts. Accordingly, the crystal structures of the isolated IF2 (11, 12) also indicate some conformational freedom of the N domain.

In the 70S IC, there is a mass of density next to the G domain reaching over to the 50S subunit (Fig. S5). Although part of this density could originate from the N domain, the remainder can be attributed to the L7/L12 protein region that extends from the L11 area of the 50S subunit into the factor binding site of the ribosome. These observations are consistent with evidence for interactions of IF2 with the L11 region (22). This region is known to have variable L7/L12 stoichiometry (23) and is rather flexible and usually invisible in ribosome structures (with the exception of complexes containing elongation factors) (24, 25). Because in the 70S IC formed with IF2-Δ69, there is no visible density attributable to the L7/L12 proteins (may be structurally disordered; Figs. 2F and S5), it can be surmised that only the N-terminal part of full-length IF2 can favor a direct interaction between IF2 and the L7/L12 region. Although this area in the 70S IC with native IF2 is not defined well enough to fit the N-terminal part of IF2 and proteins L7/L12 (Fig. S5), its sole presence indicates that this portion of IF2 probably interacts with the 50S subunit through the L7/L12 region (with a possible additional interaction through the G domain). Although the existence of an interaction between IF2 and L7/L12 has been suggested previously (26–28), the present data show that this interaction indeed occurs and is mediated by the N-terminal part of IF2. Another interesting aspect is that in the absence of the N domain (and in presence of GDPNP), IF2 binds on the 70S IC in a conformation and position virtually identical to that observed in the 30S IC (with GTP bound; Fig. 2 E and F). The C1 and C2 domains are fully visible in this complex (in contrast to the 70S IC with native IF2), and C2 reaches into the peptidyl transferase center, thus competing with the binding of a classical P/P state and resulting in a peptidyl/exit (P/E) state tRNA bound to a ratcheted 70S ribosome (a P/I state in this complex would be incompatible with the assembled 70S IC because of steric clashes). The interactions of the N domain with the L7/L12 region therefore appear crucial for the formation of a functional 70S IC in which the fMet-tRNA^fMet can be positioned properly into the P site.

The electrostatic surface of the N-terminal part of IF2 (comprising helices H1N and H2N) is characterized by a bipolar distribution with at its tip a preponderance of exposed, conserved positively charged amino acids (Fig. S2F). These residues may serve for specific interactions with ribosomal components such as the L7/L12 region located in direct proximity to the N terminus of IF2. The conformational differences between the two crystal structures observed in this part of the molecule (11, 12) suggest that this region is flexible enough to become oriented toward L7/L12 when IF2 is ribosome bound. These observations are consistent with recent findings that protein L12 is required for rapid subunit association (29, 30). In light of the above findings and considerations, experiments were carried out to obtain direct evidence for a possible role of the N-terminal portion of IF2 in the 50S recruitment to the 30S IC.

N Domain Is Involved in Ribosomal Subunit Joining.

The activities of full-length IF2 and IF2-Δ69 were analyzed by fast kinetics, revealing that the deletion mutant retains essentially full activity with respect to GTP binding and hydrolysis, fMet-tRNA^fMet binding to the 30S IC, and initiation dipeptide formation on the 70S IC (Fig. 3B; Figs. S6 and S7). However, the level of subunit joining, as monitored by light scattering experiments, was reduced by ∼50% in the presence of IF2-Δ69 compared with WT IF2 (Fig. 3A). This result is consistent with the structural observation that the N-terminal 69 residues missing in the mutant are oriented toward the 50S ribosomal subunit in the 70S IC and with the premise that they probably contribute to the stabilization of the 70S IC. An involvement of the N domain of IF2 in the interactions with the 50S subunit is consistent with previous biochemical data (31–33); however, the interaction suggested to exist between this domain and the 30S subunit cannot be seen in the cryo-EM structures presented here. A possible reason for this difference is that E. coli IF2 contains two repeats of the N domain sequence observed in T. thermophilus separated by a variable region; in E. coli, the repeat present right before the G domain (equivalent to 1–69 in T. thermophilus IF2) would be involved in subunit joining, whereas only the proximal repeat or part of the variable region could be interacting with the 30S subunit. The flexibility of the N-terminal region of IF2, suggested by the comparison of the cryo-EM structures of the 30S IC containing IF2-Δ69 or WT IF2 and by the conformational difference between the two crystal structures of IF2 (11, 12), may help sensing for the presence of the 50S subunit but may also lead to a higher degree of molecular dynamics of the 30S IC before 50S docking stabilizes the complex. The dynamic aspects of subunit joining as a function of the nature of the IF2 molecule present was then analyzed by single-molecule fluorescence.

Fig. 3.

The N domain is necessary for stable ribosomal subunit joining. (A) IF2-Δ69 shows subunit joining activity reduced by 50% compared with WT IF2, as monitored by light scattering experiments. (B) Dipeptide formation on the 70S IC. (C–F) Single-molecule subunit joining experiments with full-length or truncated T. thermophilus IF2-Δ69, E. coli IF2α, and E. coli IF2β; error bars represent SD. In C, all 50S binding events were fitted, whereas in D only the longest event for each molecule was selected for fitting. The frequencies of 50S binding events are calculated from the number of events per-trace divided by the total experiment time (E). (F) Waiting times from the delivery of 50S until the appearance of the first 50S arrival event.

Single-Molecule Experiments Reveal N Domain–Dependent Dynamics of IF2 in Subunit Joining.

To probe the efficiency of IF2 in promoting the joining of the 50S subunit to the 30S IC, we performed single-molecule fluorescence experiments with Cy3B-labeled 30S subunits and Cy5-labeled 50S subunits from E. coli (7). 30S ICs containing full-length IF2 or IF2-Δ69 were immobilized via a biotinylated mRNA, and 50S-Cy5 were delivered to start the experiment. For full-length IF2, the overall event lifetime for 50S subunit joining was 10.78 ± 0.35 s, with the longest events being 99 ± 10 s (Fig. 3 C–F). For IF2-Δ69, the 50S lifetimes are significantly shorter both when fitting all binding events (1.781 ± 0.043 s) and the longest event (16.8 ± 1.8 s). These values clearly indicates that the deletion mutant is unable to promote stable subunit joining. The high binding event frequency of the truncated IF2 at 0.080 ± 0.009 s⁻¹ (compared with 0.038 ± 0.004 s⁻¹ for full-length IF2), combined with its short 50S lifetimes, suggests that it cannot effectively guide the 50S to form stable intersubunit contacts, leading to many abortive joining attempts. As illustrated by the cryo-EM structure of the 30S IC/IF2-Δ69 complex, there is no indication that in these experiments IF2 could not be bound in the absence of the N domain.

Because the N domain of eubacterial IF2s is rather variable (Fig. S1), we wondered whether the length of this domain could have an influence on subunit joining, and we therefore performed experiments with the two isoforms of E. coli IF2. Both isoforms formed 70S complexes efficiently, with long total 50S subunit lifetimes of 23.1 ± 1.1 and 23.2 ± 1.1 s when all 50S binding events are fitted (Fig. 3 C–F). Although these lifetimes are longer than that of full-length T. thermophilus IF2, initiation factors from the same species could drive subunit joining more efficiently. The longest event of each ribosome for both isoforms was also long, at 92 ± 9 and 109 ± 11 s for IF2α and IF2β, respectively. This suggests that for both isoforms, subunit joining is generally very stable, and a significant number of 50S binding events involve stable subunit joining. The waiting times until the first 50S binding event after the start of the experiment were similar for both isoforms at 10.2 ± 1.0 and 9.9 ± 1.1 s. The frequencies of total events were low for E. coli IF2 isoforms at 0.0178 ± 0.0018 and 0.0205 ± 0.0022 s⁻¹, indicating that very few attempts are required before a successful subunit joining event occurs. In comparison, for T. thermophilus IF2-Δ69, the time until the first 50S binding event is similar to that of all other IF2s tested (11.9 ± 1.3 s), suggesting that the inability to form stable 70S complexes is not likely due to reduced 50S arrival times, which are accelerated by the presence of all of the IF2s on the 30S ICs. Taken together, the N domain length in E. coli IF2 has little influence. However, the N domain is required for stable 70S complex formation as illustrated by the T. thermophilus IF2-Δ69 mutant: its absence leads to increased but unstable sampling by the 50S subunit.

Discussion

An integrated structural biology approach, which includes high-resolution crystallographic data, solution studies (SAXS and cryo-EM), and functional analyses, enabled elucidating the full-length structure of both isolated and ribosome-bound bacterial IF2 and to attribute to its N-terminal part a specific function in ribosomal subunit joining. The precise IF2 topology derived from the present data and refined as a full-length model by MM simulations (Experimental Procedures) allows a detailed interpretation of the cryo-EM maps of the 30S and 70S ICs. The structure shows the presence of a kink around Pro355 in helix H8, close to where S12 and IF1 are bound on the 30S subunit; this proximity suggests that directly or indirectly, and without significantly affecting the thermodynamic stability of the IF2–ribosome complex (34), S12 (and part of IF1) may contribute to positioning domains C1 and C2 in such a way that interactions can be established with the fMet-tRNA^fMet on the 30S IC. The structure reveals that in the functional 30S context IF2 adopts a well-defined, extended conformation; furthermore, the C1 and C2 domains are rather flexible in solution, in the crystal structure, or on the 70S IC after release of the interactions with the fMet-tRNA^fMet (unless the N domain is removed, in which case these domains remain positioned like in the 30S IC and thereby prevent correct fMet-tRNA^fMet binding in the P site, which is consistent with reduced dipeptide formation). The observation that on the 30S IC the C1 domain interacts with residues close to the nucleotide binding site and with the switch II region suggests that nucleotide-dependent conformational changes of IF2 occurring during GTP hydrolysis (involving a reorientation of the catalytic His130 located in switch II) (11) could be transmitted directly to the C1 domain, which in turn may have an influence on the C2 domain and its interactions with the fMet-tRNA^fMet, which are released on subunit joining. Comparison of the crystal structure of isolated IF2 and the cryo-EM structure of the factor within the 30S IC demonstrates the occurrence of a remarkably large rotation of the C1 domain. This finding highlights a significant conformational flexibility of IF2 that would have been difficult to predict from a single structure alone. On the other hand, the existence of this rotation and the fact that C1 and C2 are flexible in solution (21) defy the occurrence in bacterial IF2 of large conformational changes proposed on the basis of a structural homology between IF2 and a/eIF5B (11, 20). The occurrence of a large movement of the C-terminal region with respect to the IF2 core, resulting from lever arm amplification of a relatively small nucleotide-dependent conformational change in the G domain, would rely on the existence of a rigid linker connecting the core of the molecule and its C-terminal region, which, at least in bacterial IF2, is not supported by the present data or the companion paper (12). The mechanism of action of eubacterial IF2 during translation initiation is thus unique compared with other species and also with respect to elongation factors such as EF-G, which are not involved in subunit joining.

This study also addresses the specific functions of the N and G1 domains, which display large size and sequence variability in the bacterial kingdom of IF2. In both E. coli and B. stearothermophilus the N domain provides a strong anchor for the interaction with the 30S subunit, but its presence proved to be dispensable for all IF2 functions in vitro and in vivo (31, 32, 35). A larger deletion, which included part or the totality of the G1 domain, caused a stronger phenotype in vitro (32) and in vivo because the cells bearing this deletion were able to survive at 42° C but not at 37 °C (35). The sequence alignment presented here reveals that the N-terminal region of T. thermophilus IF2 corresponds to the G1 domain in E. coli that is found duplicated at the very N terminus of the protein. Unlike the other domains of IF2, for which specific functions have been found, the function of G1 has remained thus far elusive (32). Our results show that the N-terminal region represents a key functional and structural element involved in subunit joining during translation initiation and identifies this as the region of IF2 interacting with L7/L12 of the 50S subunit. Consistently, removal of this region of the IF2 molecule results in reduced structural heterogeneity of the 30S initiation complex and increased flexibility of the L7/L12 region in the 70S IC. In addition, the L7/L12 region appears to contact the G domain of IF2; this could have direct implications on the activation of GTP hydrolysis on subunit joining (because the G domain is perfectly preorientated within the 30S IC toward the 50S) in light of the fact that the crystal structure of isolated IF2 (1–363) revealed that the catalytic His130, located within the G domain, is oriented away from the GTP ligand (11). In agreement with these observations, removal of the 69 N-terminal residues has little influence on GTP binding and hydrolysis (Figs. S6 and S7), despite the proximity between the N and G domains, but strongly impairs subunit joining activity and 70S IC stabilization as indicated by light scattering and single-molecule fluorescence data; this is consistent with the observation that the IF2-Δ69 mutant remains bound on the 70S IC in a nonproductive conformation. The latter results, which illustrate the dynamics of IF2-dependent subunit joining at the molecular level, show that in the absence of the N-terminal 69 amino acids, the lifetime event for 50S subunit joining is significantly reduced compared with full-length IF2, indicating that the presence of the N terminus of the factor favors productive sampling and 70S IC formation.

Experimental Procedures

Cloning, expression, and purification of the full-length T. thermophilus IF2 (1–571) were performed as described (3, 11, 12). SAXS data were collected at the SWING beamline at the Optimized Source of LURE Intermediary Energy Light (SOLEIL) Synchrotron (Gif-sur-Yvette, France). Cryo-EM structure determination was performed as previously described (3, 36, 37) using data collected on the in-house FEI Tecnai F30 (Polara) cryo electron microscope (SI Experimental Procedures). Sample homogeneity was tested through 3D-SC (3) using the IMAGIC suite (38), and structure refinement was done using the EMAN2 software package (39). Domain positioning of IF2 was done manually and refined through MM and flexible fitting (SI Experimental Procedures). Multiple sequence alignments were done with PipeAlign (40) (http://bips.u-strasbg.fr/PipeAlign/) and figures with PyMOL (http://pymol.sourceforge.net). Kinetics of the 30S IC–50S ribosomal subunit association were performed by stopped-flow fast kinetics (SI Experimental Procedures). Single-molecule subunit joining experiments were performed on a zero-mode-waveguide (ZMW) microscope setup (7).