Translation Initiation

Yves Mechulam,* Sylvain Blanquet, and Emmanuelle Schmitt

[SECTION EDITORS: MATHIAS SPRINGER and RICHARD BUCKINGHAM]

Posted 9 June 2011

Laboratoire de Biochimie, Ecole Polytechnique, CNRS, UMR 7654, F-91128 Palaiseau cedex, France

*Corresponding author. Mailing address: Laboratoire de Biochimie, Ecole Polytechnique, CNRS, UMR 7654, F-91128 Palaiseau cedex, France. Phone: +33 1 69 33 48 85, Fax: +33 1 69 33 49 09, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

Selection of correct start codons on mRNAs is a key step required for faithful translation of the genetic message. Such a selection occurs in a complex process, during which a translation-competent ribosome assembles, eventually having in its P site a specialized tRNA^Met base paired with the start codon on the mRNA. This chapter summarizes recent advances describing, at the molecular level, the successive steps involved in the process. In particular, structural analyses concerning complexes containing ribosomal subunits, as well as detailed kinetic studies, have shed new light on the sequence of events leading to initiation of protein synthesis in Bacteria.

OVERALL VIEW AND SYSTEM COMPONENTS

Translation initiation is a multistep process (Fig. 1). At the beginning of translation, the ribosome resulting from a previous translation round must be dissociated into 30S and 50S subunits, and the subunits must be prevented from reassociating. In addition, the system has to select the correct initiator tRNA among the pool of all cellular tRNAs and to select the correct initiation codon on mRNA. These tasks take place on the 30S subunit (169). Finally, the 50S subunit is allowed to join the initiation complex. Three protein initiation factors, named IF1, IF2, and IF3, assist the ribosome during the whole procedure. The three factors exert essential kinetic controls at various steps of the pathway, ensuring its speed and accuracy (e.g., references 6, 7, 76, 77, 78, and 92).

Fig. 1Overall scheme for initiation of translation. The involved partners are schematized using the color code indicated in the center of the figure.

Dissociation of the postterminating ribosome into its components (Fig. 1, step 1)—the 30S subunit, the 50S subunit, the last deacylated tRNA, and the mRNA—results from the combined actions of the ribosome release factor (RRF), the elongation factor G (EF-G) translocase, IF1, and IF3 (1, 94, 95, 111, 187). The whole procedure, which requires hydrolysis of an EF-G-bound GTP, can be viewed as the final step of translation termination. Therefore, EF-G and RRF are usually not considered as initiation factors. The exact sequence of events depends on the mRNA studied. In some cases, IF3, together with IF1, is involved in the dissociation of deacylated tRNA from the 30S subunit, just after 50S subunit dissociation (111, 190). Other reports favor the idea that the deacylated tRNA is ejected before subunit dissociation (1, 95). A recent study has proposed that the two RRF/EF-G and IF1/IF3 pathways of dissociation coexist and act with different efficiencies, depending on the strength of mRNA binding to the ribosome. The RRF/EF-G pathway would constitute the major route operating on posttermination ribosomes after peptide release, while the IF1/IF3 pathway may operate on posttermination ribosomes that lost tRNA from the P site (189). This IF1/IF3 pathway would be efficient to activated stored ribosomes (i.e., resulting from the activation of superfluous ribosomes that had been stored as inactive complexes) in the case of upshift in the growth medium or of peptidyl-tRNA drop-off during steady-state growth (189).

Once dissociation has occurred, it is clear that IF3 plays a crucial role in maintaining 30S ribosomes available for initiation. IF3 impairs the reassociation of ribosomal subunits through binding to the free 30S subunit, thereby decreasing the ribosomal subunit association rate. IF1 stimulates this activity of IF3 (69). Finally, the small ribosomal subunit recruits the three initiation factors, IF1, IF3, and IF2 bound to a GTP molecule, as well as an mRNA and the initiator tRNA charged with formylmethionine (Fig. 1, steps 2 and 3). Binding of the mRNA is most often guided by base pairing between the Shine-Dalgarno (SD) sequence, located a few nucleotides upstream from the AUG start codon, and the anti-Shine-Dalgarno (aSD) sequence at the 3′ end of 16S RNA (240). Then, a series of conformational rearrangements occur (273), leading to a stable 30S initiation complex (30SIC) with proper AUG initiation codon/CAU tRNA anticodon base pairing (Fig. 1, step 4). The kinetic control exerted by the initiation factors during these rearrangements is crucial for accuracy (6, 21). The 30SIC then recruits the 50S subunit, while IF3 and IF1 are released. This process, called subunit joining (Fig. 1, step 5), is catalyzed by IF2. During step 5 (Fig. 1), energy is consumed through the rapid hydrolysis of IF2-bound GTP into GDP and P_i. Then, P_i is released and IF2 dissociates in a GDP-bound form (5, 71). This yields an active 70S initiation complex (70SIC) with initiator tRNA in its P site, base paired with the initiation codon. The elongation phase can subsequently begin.

Although the initiation mechanism outlined above is by far the most widely used in eubacteria, some cases of leaderless mRNAs have been reported (163). In these mRNAs, the start AUG codon is located at or very close to the 5′ end. In gram-negative bacteria, such leaderless mRNAs are quite rare; the best-documented case is that of the bacteriophage λ cI gene (9). Initiation of translation of leaderless mRNAs escapes the canonical scheme and will be discussed at the end of this chapter.

STRUCTURE OF THE 30S RIBOSOMAL SUBUNIT

The Escherichia coli small ribosomal subunit is composed of a 16S rRNA (1,542 nucleotides [19]) and of 21 proteins. Following early images of 30S subunits obtained by electron microscopy (125, 260, 269), progress in cryo-electron microscopy techniques (cryo-EM) led to a precise view of the ribonucleoprotein particle (62, 128). In parallel, the efforts to obtain high-resolution diffracting crystals of the Thermus thermophilus 30S subunits (263, 277, 278, 280) led to high-resolution structures in the year 2000 (224, 272). The classical description of the subunit uses morphological elements known as the “head” (Fig. 2) and the “body” connected through the “neck.” Three elements—the “platform,” the “shoulder,” and the “spur”—can be distinguished within the body. This overall shape is mainly due to the 16S rRNA 3D scaffold, with each morphological element roughly corresponding to one domain in the nucleic acid. The body is made of the 5′ domain, while the central domain and the 3′ major domain correspond to the platform and the head, respectively. The fourth (3′ minor) domain folds back on the body, at the interface with the 50S subunit. These four domains are centered on the neck region, near the decoding center where mRNA interacts with tRNA (283). Notably, crystal structures of the E. coli 70S ribosome have revealed high flexibility of the head region, which can rotate around the neck. Such a flexibility is likely to be involved in tRNA and mRNA movements during translation (233, 284).

Fig. 2Schematic views of the 30S subunits. (A) Ribbon representation of the 16S rRNA. In the three panels, the four domains are colored as follows: 5′ domain, red; central domain, green; 3′ major domain, yellow; 3′ minor domain, blue. Morphological elements are indicated. (B) Same as panel A, but the ribosomal proteins are shown as grey ribbons. Panels A and B were drawn by using the coordinates of the 30S subunit in the PDB file 1J5E (272). (C) Position of the three tRNA sites. The 30S subunit is represented as a surface, using the same orientation as in panels A and B. The three tRNAs are colored as follows: A-site tRNA, red; P-site tRNA, blue; E-site tRNA, cyan. Panel C was drawn by using the coordinates in the PDB file 2HGP (282). Figure 2 was drawn with PyMol (DeLano Scientific LLC).

mRNA BINDING AND ACCOMMODATION

Statistical analysis of the nucleotide sequences of translation initiation regions (TIRs) in the genome of E. coli indicated nonrandomness of bases between position –20 and position +13, with position 0 being the A in the AUG initiation codon (229). This nonrandom region corresponds to that protected by an initiating ribosome. It was called ribosome-binding site (RBS). Two spots in the RBS are highly biased: the AUG initiation codon itself, with GUG and UUG also encountered, and the SD sequence (having a consensus UAAGGAGGU). Another notable bias occurs at position 3, where A is preferred over C, U, or G (in this order [11]). Random search for TIRs active in E. coli using a library made from eukaryotic noncoding DNA showed that the above bias may indeed be part of the initiation signals (47). Deviations from the statistical consensus were observed to systematically decrease the translation initiation rate (89). Moreover, a spacing of 5 nucleotides between the last U in the SD consensus and the start codon was experimentally shown to be optimal for the rate of gene expression (27, 89, 209).

The SD consensus shows exact complementarity with the 3′ end of the 16S rRNA (240). This 16S rRNA region was therefore termed an aSD sequence. Crystal structures of bacterial ribosomes with RNA fragments containing an SD sequence allowed direct visualization of the SD-aSD duplex, sitting on the platform of the small ribosomal subunit (109, 283). The importance of forming the SD-aSD helix for anchoring the mRNA to the ribosome is underscored by the requirement of free SD accessibility for efficient translation (15, 21, 152, 208). Moreover, quantitative analysis of the relationships between rate of translation and mRNA secondary structure has led to the suggestion that initiation depends on spontaneous unfolding of the entire initiation region (40).

However, many examples demonstrate that a 30S ribosomal subunit is able to bind structured mRNAs (e.g., 106, 177, 208, 214, 283). In such cases, strong SD-aSD pairing helps the ribosome to compete with mRNA intramolecular pairing (41). Because of the rapid folding kinetics of mRNA, such a competition model implies that the 30S subunit stays in a standby site, in contact with the mRNA while it is still folded, ready to recognize the SD sequence as soon as the structure opens (38, 42). This second step leads to formation of the SD-aSD duplex (90, 251). The first step of binding at the standby site is independent of the length of the SD sequence (251). Rather, the available data indicate that it is driven by contacts between ribosomal protein S1 and AU-rich and/or pseudoknot-containing single-stranded mRNA sequences upstream from the SD sequence (17, 46, 98, 117, 207, 244). Such S1-driven contacts may even substitute for an SD sequence, in cases where the latter is weak or absent (116, 243). After this first, fast, binding phase, unfolding of the mRNA occurs (42, 251). Interestingly, this step appears much accelerated in the simultaneous presence of the SD, the start codon, the initiator tRNA, and the GTP-bound form of IF2, thereby expressing an unexpected reciprocal link between the SD-aSD and the initiator tRNA-start codon pairings (251). The binding of slightly structured mRNAs to the ribosome also follows a multistep process. In this case, however, the initial binding step would more often directly involve the SD-aSD pairing (23, 122).

Several X-ray and cryo-EM structures have allowed direct visualization of mRNAs at various stages of binding and adaptation. A translational control mechanism, called entrapment, is based on the possibility for the regulatory molecule to block the accommodation of a structured mRNA (212, 223). This mechanism is illustrated by the control exerted by ribosomal protein S15 on translation initiation of its own mRNA (54, 196). The cryo-EM structure of an initiation complex blocked in a preaccommodation state by S15 binding was obtained by cryo-EM (149). The folded mRNA sits on the platform of the ribosome, in the vicinity of conserved amino acid residues in ribosomal proteins S2, S7, S11, and S18. Interestingly, the same region of the ribosome is involved in the docking of other structured mRNAs (106). Moreover, stable stem and loop structures formed by poly(A)- or poly(U)-rich regions upstream from the SD were also observed sitting on the platform of the ribosome (282). These observations suggest the existence of a common docking site on the platform of the 30S subunit for the initial binding of mRNAs (241). A further intermediary state in mRNA-30S subunit association was revealed, with the SD-aSD helix stabilized in a chamber between the head and the platform of the ribosome, before that slight motion of the helix occurs in response to initiator tRNA-start codon pairing (109). Once completely accommodated, the mRNA is wrapped in a groove that encircles the neck of the 30S subunit (282, 283).

Recently, a new function has been evidenced for the SD-aSD interaction. This function is expressed beyond anchoring mRNA to the ribosome. Indeed, occupation of the tRNA E site contributes to avoid erroneous selection of noncognate aminoacyl-tRNAs at the A site (67). Just after the initiation phase the E site is still empty, which might cause errors during incorporation of the second amino acid. However, the presence of the SD sequence was shown to functionally compensate for the absence of tRNA in the E site, thus conferring accuracy to the decoding step following initiator tRNA selection (43).

Section summary

Several features cooperate to regulate the rate of initiation of translation of an mRNA. These features include the initiation codon, the SD sequence and its position with respect to the start codon, upstream enhancer sequences, and the occurrence of secondary structures on the mRNA. Binding and accommodation of the mRNA is a multistep process, some steps of which depend on the initiation factors and on the pairing of the initiator tRNA to the start codon on the mRNA. Indeed, whatever the mRNA, it is finally the base pairing between the anticodon of the initiator tRNA and the start codon of the mRNA that will precisely position the initiation complex and set the correct reading frame.

INITIATOR IDENTITY OF tRNAfMet

Because of the utmost importance of the initiator tRNA in setting the correct reading frame on the mRNA, selection of this tRNA by the initiation machinery is a key event. Specific functional features of the initiator tRNA ensure its intake by the initiation machinery (binding to IF2 and to the P site of the 30S ribosome), its exclusion by elongation factor Tu (EF-Tu), and its resistance to the action of peptidyl-tRNA hydrolase (52, 119, 158, 230). The initiator tRNA universally is a methionine tRNA, and the insertion of methionine as first amino acid may be considered as part of the translation initiation signal (reviewed in references 153 and 203). Aminoacylation by methionyl-tRNA synthetase is mainly governed by the CAU anticodon (155, 227, 231, 232), although other, less stringent, determinants exist in the acceptor stem (154) (Fig. 3). E. coli K-12 contains two tRNA_f^Met species, differing only by one nucleotide at position 46 in the extra arm, which can be either 7-methyl-G (tRNA_f1^Met) or A (tRNA_f2^Met) (32, 48, 49, 53). In contrast, E. coli B contains tRNA_f1^Met only (144). The reason for the occurrence of these two tRNA species is not understood.

Formylation of the Initiator Methionine

In Bacteria, as well as in mitochondria and chloroplasts, the methionine esterified to the 3′-terminal adenosine is systematically N-formylated, thanks to the action of methionyl-tRNA_f^Met formyltransferase (MTF) (30, 31, 147). Formylation plays a dual role in vivo, both favoring the use of tRNA_f^Met in initiation and disfavoring its use in elongation of translation (83, 84, 145, 235, 266, 267, 268). Accordingly, disruption of the gene encoding MTF is highly detrimental to the growth of E. coli  since it results in an increase of the doubling from 26 min to 215 min at 37°C and in the absence of growth at 42°C (82). Several lines of evidence, obtained in vitro as well as in vivo, indicate that the importance of the formyl group is mostly linked to its role in initiator tRNA binding to IF2 (124). In particular, N-blocking of Met-tRNA_f^Met strongly increases the affinity of the nucleic acid for IF2 (193, 253) and favors its selection by IF2 in the P site of initiating 30S subunits (88). In line with such a role, overproduction of IF2 partially restores the initiator activity of unformylated mutant tRNAs in vivo (83, 150). Finally, the less stringent sensitivity of Pseudomonas aeruginosa to disruption of its formylase gene is due to specific properties of IF2 in this organism. These features were proposed to allow the factor to significantly bind unformylated initiator tRNA (173, 248).

Formylase specifically catalyzes the transfer of a formyl group from N-10-formyltetrahydrofolate to the α-amino group of methionine esterified on tRNA_f^Met (44, 108). Determinants in the sequence of the polynucleotidic substrate govern the efficiency of the formylation reaction. In E. coli, these determinants (Fig. 3) were shown to be (i) the top of the acceptor stem, (ii) the methionine side chain, and (iii) the A11-U24 base pair (68, 85, 133). Among these, the most important feature is the presence of mismatched bases, C1-A72, at the top of the acceptor stem, while most elongator tRNAs have GC or CG at this position. Notably, bases 1 and 72 are mismatched in all bacterial initiator tRNAs (146). The discriminator base, A73, as well as the G2-C71, C3-G70, and G4-C69 pairs, also markedly influences the E. coli formylation reaction (85, 132, 133). Compared with the role of the acceptor stem, the presence of a purine-pyrimidine pair at position 11–24, also characteristic of initiator tRNAs, has a modest effect on the efficiency of the formylation reaction (133). Finally, a methionyl moiety at the 3′ end of the tRNA is the preferred one, although other amino acids can be formylated provided that they are esterified on an initiator tRNA (68, 85, 136). Notably, a methionine side chain in an esterified initiator tRNA is also important, although not strictly required, for IF2 binding (136, 150, 275).

Fig. 3Roles of the determinants of tRNA_f^Met important for its use in initiation. The sequence of E. coli tRNA_f2^Met is shown in the cloverleaf representation. Important parts are boxed using the color code indicated on the figure.

Determination of the crystal structures of E. coli formylase (225) and of its complex with initiator tRNA (228) has provided the molecular basis explaining tRNA recognition (Fig. 4). The enzyme is made of a Rossmann-fold domain containing the catalytic center, connected to a C-terminal β-barrel by an extended linker. An enzyme loop (loop 1 in Fig. 3) in the active site domain is wedged in the major groove of the acceptor stem of the tRNA, thereby splitting the C1 and A72 bases and bending the 3′ arm toward the active center. The bent conformation of the acceptor arm would not be possible with paired 1–72 bases, thereby explaining why GC or CG base pairs at position 1–72 are highly detrimental to the formylation reaction. Lysine and arginine residues of the enzyme loop make base-specific contacts with the acceptor stem. The OB-fold domain binds the D stem through a base-specific contact with U24.

Fig. 4Schematic representation of methionyl-tRNA_f^Met transformylase from E. coli complexed with its product formyl-methionyl-tRNA_f^Met. The N-terminal domain of the enzyme is green, and its C-terminal domain is yellow. Bases 1 and 72, at the top of the acceptor stem of tRNA, are separated by loop 1 protruding out of the enzyme. The figure was drawn using the coordinates in the PDB file 2FMT (228). Figure 4 was drawn with PyMol (DeLano Scientific LLC).

Rejection by EF-Tu

Apart from its role in directing interaction with IF2, the formyl group prevents interaction of the initiator tRNA with EF-Tu (105, 256). The origin of this rejection is clearly explained by the structure of the ternary complex EF-Tu:GTP:Phe-tRNA^Phe, showing a close interaction of the enzyme with the amino group of the esterified amino acid (175). Notably, the cellular concentration of EF-Tu is much larger than that of MTF, and in vivo studies have shown that EF-Tu and MTF can compete for the uptake of Met-tRNA_f^Met (83). However, even before formylation, binding of Met-tRNA_f^Met to EF-Tu is disfavored in comparison with that of an aminoacylated elongator tRNA (105). Such a behavior may again be linked to the C1-A72 mismatch. Indeed, EF-Tu tightly binds the 5′-phosphoryl group of its aminoacyl-tRNA ligands (175, 176). Possibly, a 1–72 mismatch renders the phosphoryl group more mobile. As a result, the binding of methionylated initiator tRNA to EF-Tu would be weakened. This mechanism was proposed by analogy with that explaining the resistance of fMet-tRNA_f^Met to the action of peptidyl-tRNA hydrolase (see below). Moreover, EF-Tu binds tRNAs on the T-stem side, while the formylase binds initiator tRNA on the D-stem side. This led to the proposal that the formylase may bind and process Met-tRNA_f^Met, even if this molecule has already been misappropriated by EF-Tu. In this case, formylation would be possible without prior full dissociation of the erroneous EF-Tu:tRNA_f^Met complex (228). Such a reaction would clearly diminish the negative impact of the competition for Met-tRNA_f^Met between the two proteins in vivo.

Resistance to the Action of Peptidyl-tRNA Hydrolase

As a consequence of its formylation, the initiator tRNA is a potential substrate of peptidyl-tRNA hydrolase (PTH). This enzyme, essential to E. coli growth, recycles peptidyl-tRNAs resulting from aborted translation into free tRNAs (8, 65). Indeed, accumulation of peptidyl-tRNAs is toxic, probably because of the sequestration of essential tRNA species (157). More generally, PTH is active on N-blocked aminoacyl-tRNAs. However, fMet-tRNA_f^Met is uniquely protected from a futile hydrolysis by PTH. Interestingly, such a protection is again due to the mispairing of the 1–72 bases (52, 132, 230). PTH features at its surface a cationic site for the 5′ phosphate of the tRNA (61, 226). Anchoring of the 5′ end of the tRNA is thought to direct the peptidylated 3′ end of the tRNA toward the catalytic center. By changing the position of the phosphate group, the 1–72 mismatch in the initiator tRNA would impair simultaneous docking of the 5′ and 3′ ends of the tRNA on the hydrolase.

Features in the Anticodon Arm

The lack of Watson-Crick base pair at position 1–72 has a key importance in formylase recognition, rejection by EF-Tu, and rejection by PTH. However, this feature is not sufficient to fully specify the initiator identity of a tRNA in translation. Also required are the three consecutive G-C base pairs in the anticodon stem, unique to bacterial tRNAs (84, 87, 236). These three base pairs are involved in the selection exerted by IF3 at the P site on the 30S subunit. Notably, these G-C base pairs have been associated with a peculiar conformation of the anticodon loop (236, 274). In keeping with this idea, the crystal structures of tRNA_f^Met, free (10) or complexed with the formylase (228), show that base A37 is flipped out outside the anticodon loop. This will be discussed further below (see "Initiation Factor 3," below).

Section summary

The initiator tRNA, universally aminoacylated by a methionine, displays several features important for its initiator identity. These features include the absence of base pairing at the top of the acceptor stem and the presence of three consecutive G-C pairs in the anticodon stem. They account for the specific formylation of the esterified methionine and the selection exerted by IF3 in the P site of the 30S ribosomal subunit. In turn, formylation allows the initiator tRNA to bind IF2 and to be rejected by EF-Tu. Finally, the absence of base pairing at the top of the acceptor stem allows fMet-tRNA_f^Met to escape the action of peptidyl-tRNA hydrolase.

INITIATION FACTORS

A time of ca. 3 s has been estimated to be necessary to achieve initiation of protein synthesis in vivo (112). The presence of the three initiation factors (IF1, IF2, and IF3) does not markedly change the affinity of the initiator tRNA for the mRNA-programmed 30S subunit (dissociation constant of 3.6 nM in the absence of the three factors and of 2.7 nM in their presence [7]). Therefore, the factors govern formation of the initiation complex by modulating kinetic constants rather than equilibrium constants.

Initiation Factor 2

Structure of IF2

Translation initiation factor 2 (IF2) (97 kDa) is the largest of the three initiation factors. It is a G protein, able to bind GTP and to hydrolyze it into GDP and P_i during the initiation process. After initiation, IF2 is released in a GDP-bound form. Like IF1, IF2 is a universal factor since homologues are found in both eukaryotic and archaeal cells (29, 134). In E. coli, two forms, IF2α and IF2β, can be expressed. They arise from translation initiation at in-frame start codons of the infB gene (99, 198, 217). IF2α contains 890 residues (97.3 kDa). It was later observed that IF2β actually corresponds to two isoforms, IF2β and IF2β′, beginning at codons 158 and 165 (79.7 and 78.8 kDa, respectively [180]). Multiple isoforms of IF2 have also been observed in many Enterobacteriaceae (100) as well as in Bacillus subtilis (103). Although the reason for the presence of multiple isoforms of the factor is still unknown, both IF2α and IF2β are required for optimal growth of E. coli (219).

On the other hand, a minimal form of IF2 (55 kDa) resulting from the removal of the 388 N-terminal amino acids is sufficient to support growth (124). The N terminus of this minimal form corresponds to that of the archaeal IF2 homologue aIF5B. The initiation factor aIF5B from Methanobacterium thermoautotrophicum is the only IF2 homologue whose high-resolution structure is known (211). The dispensable N-terminal region of IF2-α appears to ensure optimal interaction of IF2 with the 30S and 50S ribosomal subunits (166, 167). It was divided into three domains, N1 (1–157), N2 (158–289), and N3 (290–390), with N1 corresponding to the domain lacking in IF2-β (168). The structure of the dispensable N-terminal region has been addressed by using nuclear magnetic resonance (NMR) and circular dichroism (CD). N1 displays a structured region (residues 2–50), folded as three α-helices packed on one side of a short three-stranded β-sheet, followed by a probably unstructured region (residues 61–97) and a hydrophilic α-helix (residues 98–157; Fig. 5 [130]). This N1 domain appears connected to the rest of IF2 by a flexible linker. Moreover, domains N2 and N3 contain unstructured regions and have a substantial helical content (129).

The core part of the factor, as deduced from the 3D structure of archaeal aIF5B, is made of four domains, called I (or G), II, III, and IV (Fig. 5 [211] ). The overall shape of the protein is that of a chalice, with domains I, II, and III forming the cup, connected by a long α-helix to domain IV forming the base. Domain I is a classical G domain (270), resembling the G domains of EF-Tu and EF-G. Accordingly, this domain contains the GTP/GDP binding site. Domain II is a β-barrel formed of 11 antiparallel strands, structurally similar to domains II of EF-Tu and EF-G. Domain III is made of a four-stranded β-sheet, flanked on both sides by two α-helices. Finally, the C-terminal domain IV, connected to domain III by a 40-Å-long α-helix, is an eight-stranded β-barrel, followed by two α-helices. According to sequence alignments and to the NMR structure of the C-terminal domain of Bacillus stearothermophilus IF2 (159), these two terminal α-helices are specific to e/aIF5B. They are absent in IF2. Notably, several lines of evidence indicate some mobility of domain IV regarding the rest of the structure, which may be important for IF2 functions during the translation initiation cycle (3, 204, 211, 242). The structure of domain III of B. stearothermophilus IF2 also resembles that of aIF5B, although some differences have been noticed (271).

Fig. 5Structures of bacterial (left) and eukaryotic or archaeal (right) initiation factors. The PDB files used to draw the cartoon are indicated below each picture. In some cases, the PDB files of corresponding homologues are also indicated. Topologies of IF3 domains, eIF1, IF1, and eIF1A, are indicated. Figure 5 was drawn with PyMol (DeLano Scientific LLC).

Early studies on the function of IF2

The existence of soluble factors (IF1 and IF2) able to stimulate initiator tRNA binding to ribosomes in the presence of GTP in vitro was shown early (4, 131, 184, 185, 221, 247). However, it was then shown that hydrolysis of GTP into GDP and P_i was not essential to obtain initiator tRNA binding or 50S subunit joining but was required for formation of the first peptidic bond (93, 115, 185, 259). The corresponding GTPase activity was clearly attributed to IF2 (114). Further purification and characterization showed that IF2 was a large polypeptide, sharing some properties in common with EF-Tu and EF-G (113). The idea of stimulation by IF2 of initiator tRNA binding to the 30S subunit was reinforced by the identification of a complex between IF2 and the initiator tRNA (137). These early studies also led to the view that, in addition to the stimulation of initiator tRNA binding to the 30S subunit, IF2 favored the association of the 50S subunit to the 30S initiation complex (50, 69). Hydrolysis of GTP by IF2 was proposed to “unblock” the initiator tRNA for peptidyl transfer and to trigger release of IF2 in a GDP-bound form (50, 138). Since these pioneering studies, a large amount of work has led to a more precise description at the molecular level.

Mode of action and nucleotide cycle of IF2

Binding of initiator tRNA to IF2 occurs with a dissociation constant in the micromolar range. It does not depend on the presence of GTP or GDP (193, 253). This behavior is markedly different from that of EF-Tu, to which aminoacyl-tRNAs bind with dissociation constants in the nanomolar range, in a strictly GTP-dependent manner (105). Moreover, unlike EF-Tu, which utilizes all of its three domains to bind tRNA (175), domain IV of IF2 is sufficient for initiator tRNA binding (79, 246). Interaction studies performed with B. stearothermophilus or T. thermophilus IF2 showed that the formyl-methionylated 3′-terminal hexanucleotide of tRNA contains all the thermodynamic determinants governing interaction and that fMet-AMP or even fMet behaves as a minimal but specific ligand (81, 246, 255). This behavior agrees with the observations that recognition of the initiator tRNA is mainly based on the presence of an N-blocked methionine esterified to tRNA (136, 150, 193, 253). However, IF2 is also sensitive to the absence in the initiator tRNA of strong base pairing at position 1–72 (150). Because of the relatively small affinity of IF2 for the initiator tRNA, a role of tRNA carrier, analogous to that of EF-Tu, is unlikely for IF2. Most probably, the normal mode of action of IF2 is to bind the 30S subunit first and then recruit the initiator tRNA (7, 22, 78, 92). However, under extreme conditions, such as IF2 overproduction, the occurrence of an IF2-initiator tRNA complex outside the ribosome may have some significance in vivo (83, 275). Whatever the order of binding between 30S subunit, IF2, and tRNA, it is clear that, in addition to promoting the association of ribosomal subunits, a key role of IF2 is to enhance the binding of the initiator tRNA by increasing its association rate with the small ribosome subunit (7, 72). It is important to emphasize that the promoting action of IF2 on formation of the 70S complex is selective of the initiator tRNA through recognition of its formyl group. IF1 acts synergistically to assist IF2 in promoting formation of this complex ([6]; see “Role of IF1 in the rate of initiation of protein synthesis and accuracy,” below).

Initiator tRNA is better stabilized on the 30S subunit by the GTP form of 30S-bound IF2 than by the GDP-bound form (5). Moreover, the GTP-bound form of IF2 promotes faster joining of ribosomal subunits than the GDP-bound form does (5, 71). This behavior agrees with the idea that hydrolysis of GTP by IF2 drives a series of rearrangements within the initiation complex. These rearrangements are in turn necessary for the final release of IF2 from the 70S initiation complex in its GDP-bound form (7, 71, 142). More precisely, the nucleotide cycle of IF2 includes a fast hydrolysis of GTP into GDP and P_i. This step is triggered upon the joining of the 50S ribosomal subunit to the 30S initiation complex (7, 71). A further step is delineated by the slow release of P_i (71, 262). Notably, P_i release from the eukaryotic initiation factor eIF2 was also identified as a key event during start codon selection (2, 28). This view is further supported by structural data obtained with the archaeal homologue of eIF2 (276).

Finally, it is notable that IF2 orthologues are found in all cells, including archaea and eukaryotes, where they have been named aIF5B and eIF5B, respectively (29, 121). In eukaryotes and in archaea, initiator Met-tRNA_i^Met is carried toward the ribosome by a heterotrimeric e/aIF2 factor. The occurrence of this factor, which has no orthologue in Bacteria, raised the question of the utility of the e/aIF5B factor. However, several lines of evidence have shown that e/aIF5B plays a role similar to that of IF2 in initiation of translation (5). In particular, GTP-bound eIF5B promotes ribosomal subunit joining (192, 265), and GTP hydrolysis is required for release of eIF5B from the 80S ribosome (135, 239). Furthermore, genetic evidence has supported the idea that eIF5B facilitates the binding of initiator tRNA to the small ribosomal subunit (29), and e/aIF5B has been shown to bind initiator Met-tRNA_i^Met in vitro (86). Hence, IF2 and e/aIF5B have kept some common functions throughout evolution, including verification of the presence of a correctly positioned initiator tRNA in the ribosomal P site before the two ribosomal subunits join together. IF2 and e/aIF5B would therefore be engaged in a final checkpoint before translation can begin (3, 239).

Section summary

There is a consensus in the current literature that IF2 facilitates (i) initiator tRNA binding, (ii) ribosomal subunit joining, and (iii) adjustment of the initiator tRNA in the P site of the ribosome. Motions of IF2 occur on the ribosome during these actions. The four available cryo-EM structures of IF2 bound to the ribosome, thought to illustrate different states, support the idea of such motions (3, 171, 242; see below).

Initiation Factor 1

Translation initiation factor 1 (IF1) is a small RNA-binding protein (71 amino acids in E.coli) made up of a five-stranded β-barrel containing a short capping α-helix between strands 3 and 4 (237). It is encoded by the gene infA (222), found in all prokaryotic genomes and in chloroplasts (96), but not in mammalian mitochondria, where its function is taken over by a particular IF2 (66). IF1 is essential to cell viability in E. coli (35).

Site of binding of IF1 to the 30S ribosomal subunit

The site of binding of IF1 to the 30S ribosomal subunit has been extensively studied, both at the biochemical and structural levels. Specific protection of A1492 and A1493 in 16S rRNA from modification by dimethyl sulfate and protection of G530 from kethoxal were observed upon binding of the factor to 30S (96, 162). Moreover, mutational analysis showed that the bases C1407, A1408, A1492, and A1493 make up part of the IF1 binding site (36). On the other hand, site-directed mutagenesis studies have led to the identification of IF1 residues important for the function of this factor (33, 34, 80). In 2001, the resolution of the crystallographic structure of a T. thermophilus IF1:30S complex allowed a fine description of the binding mode of the factor to the ribosomal subunit (24). This structure is in keeping with the biochemical results previously obtained. IF1 binds to the 30S subunit in a crevice delineated by helix H44 and the G530 loop of the 16S rRNA, as well as by protein S12. As a consequence, IF1 blocks the A site (Fig. 6A). Moreover, the binding induces a flipping out of the A1492 and A1493 bases from helix H44 toward a pocket of IF1 delineated by the loop 17–25. This sequestering explains why IF1 protects these two bases from chemical modification (36, 162). Comparison of the crystallographic structure of the IF1:30S complex (24) with that of isolated 30S subunit crystallized under the same conditions (272) shows that, beyond localized changes around the A site, the binding of IF1 also causes long-distance conformational changes in H44, which is slightly displaced over a distance of 70 Å. Small movements of the domains of the 30S subunit with the head, platform, and shoulder all rotating toward the A site are also visible. These movements are possibly underestimated, since they may have been limited by crystal-packing constraints (24).

Role of IF1 in the rate and accuracy of initiation of protein synthesis

The reason why IF1 occupies the A site during initiation of translation is still under study, although it is obvious that, at this binding position, the factor inhibits the binding of any tRNA in the A site. Surprisingly, IF1 alone has no effect on the rate of ternary complex formation (30S:N-ac-tRNAPhe-mRNA_PolyU [273] or 30S:fmet-tRNA_f^Met:mRNA_MFT1 [7]). Only a small effect (2.5-fold) was observed when poly(AUG) was used as the mRNA (200). The presence of IF3 or of IF2 was necessary to detect an effect of IF1 on the kinetics of 30S preinitiation complex formation (7, 273). Indeed, on one hand, it was observed that IF1 amplifies the destabilizing effect of IF3 on the binding of fmet-tRNA_f^Met to a 30S:mRNA complex. On the other hand, in the absence of IF3, the addition of IF1 amplified the stabilizing effect of IF2 on the binding of fmet-tRNA_f^Met to the 30S:mRNA complex (7). In keeping with these observations, IF1 specifically prevents the dissociation of IF2 from 30S complexes in a manner independent of IF3 (250). Such a role is important since it ensures that IF2 can reach the step of GTP hydrolysis. This step facilitates the formation of an active elongation-competent 70S complex.

Upon using noninitiator tRNA, an amplification by IF1 of the destabilizing effect of IF3 was also observed (6). As a consequence, removal of IF1 from the full set of initiation factors results in a nearly 60-fold reduction in the preference for initiator tRNA over elongator Phe-tRNA^Phe (6). Such behavior is sufficient to explain why IF1 is essential to cell viability (35).

Notably, like IF2, IF1 possesses an orthologue in eukaryotes and archaea named e/aIF1A. Eukaryotic eIF1A has the same OB-fold (oligonucleotide-oligosaccharide binding fold, featuring a β-barrel) core as IF1 but contains an additional helical C-terminal subdomain as well as long unstructured N-terminal and C-terminal tails (12). Together with another eukaryotic/archaeal initiation factor named eIF1, eIF1A induces a conformational change of the 40S ribosomal subunit that, in turn, triggers rapid binding of the initiator tRNA complexed with its carrier eIF2 (188). Similarly to IF1, the core of eIF1A binds the small ribosomal subunit at the A site (188, 279). However, the mode of action of eIF1A may be more complex than that of IF1, since the specific N- and C-terminal extensions of the eukaryotic factor are essential to its roles in promoting initiator tRNA-eIF2 binding to the initiation complex and in favoring faithful start codon recognition (57, 161, 220). The available data have led to the idea that the N-terminal tail is involved in promoting a close conformation of the small ribosomal subunit, while the C-terminal tail would be ejected from the P site upon start codon recognition, thereby allowing the initiator tRNA to fully enter the P site when the start codon is correct (141, 279).

Effect of IF1 on the dissociation of 70S ribosomes

IF1 can bind to 30S or 70S subunits. In the presence of IF1 at a low concentration (about 100 nM), the forward and backward rate constants of the 70S ribosome-subunit equilibrium are slightly increased. In the presence of IF3 (at a concentration of 100 nM), the association rate is decreased by a factor of nearly 5, whereas the dissociation rate is not affected. Therefore, IF1 contributes to the IF3-induced dissociation of ribosomal subunits (69, 172). Recent work, however, has shown that IF1 displays weak affinity for the 70S subunit. Indeed, the K_m of IF1 in the splitting reaction of 70S into 50S and 30S catalyzed by IF3 and IF1 is equal to 20 μM (189), while the dissociation constant for IF1 binding to the free 30S subunits is about 1 μM (36, 285). Taking into account this large K_m value, Pavlov et al. (189) showed that IF1 alone, at high concentrations, was able to induce dissociation of vacant 70S ribosomes and of 70S:mRNA complexes, whereas it did not affect the rate constant for association of the subunits. Moreover, these dissociation rates in the presence of IF1 alone are further increased upon addition of IF3. Therefore, in the presence of IF1, IF3 behaves also as a dissociating factor. These results led to the model that, upon binding of IF1 to the 70S subunit (vacant or complexed to mRNA), a “half-opened” conformation of the ribosome is induced. In this conformation, access by IF3 to its binding site on the 30S subunit would be possible. Upon this binding of IF3, the two subunits cannot close again and dissociation is rapidly obtained. Such an activity would be useful in vivo to recycle ribosomes after peptidyl-tRNA drop-off during steady-state growth or in the case of reactivation of the ribosomes during transition from the stationary to the exponential growth phase.

This model suggests that conformational changes in the 30S subunit may occur within the 70S ribosomes, leading to a conformer of the 30S moiety with low affinity for IF1. In this view, it is notable that, in the structure of the IF1:30S complex, observed conformational changes of H44 may break some of the intersubunit bridges (24). This observation may explain why, in the presence of IF1, 70S subunits can be more easily dissociated. Another recent study suggests that, by negatively regulating the formation of the 70S initiation complex, IF1 helps IF3 in the enhancement of start codon selection (160). In this view, the IF1-induced conformational change of the 30S subunit would become favorable to initiation, only after the reversion of this conformational change by cognate codon-anticodon interaction controlled by IF3 (160, 202).

Does IF1 interact directly with the two other initiation factors?

Affinity of IF1 for the 30S subunit is increased in the presence of IF3 or in the presence of IF2 (25, 26, 166, 285). These effects can result either from direct interactions between the factors on the 30S subunit or from an influence of the factors on the conformation of the ribosome. Cross-linking experiments argue in favor of a direct interaction between IF1 and IF3 on one hand and between IF1 and IF2 on the other hand (16). Although the position of the IF3-C terminal domain on the 30S subunit remains controversial, the available structural data do not favor a direct interaction between IF3 and IF1 (24; see "Initiation Factor 3," below ). Moreover, clear biochemical evidence in favor of a direct interaction between IF1 and IF2 is still lacking (110). Interestingly, it is possible to form a complex between e/aIF1A and e/aIF5B, the eukaryotic and archaeal homologues of IF1 and IF2, respectively. This complex may designate the IF1-IF2 interaction as a universal feature of translation initiation. However, the regions in e/aIF1A and e/aIF5B involved in complex formation are not conserved in the bacterial homologues.

Section summary

IF1 binds the 16S rRNA in the A site. Such a binding causes long-distance conformational changes in the 30S subunit, resulting in turn in a rotation of the head, platform, and shoulder. Synergistically with IF2, IF1 facilitates initiator tRNA binding to the initiation complex and, synergistically with IF3, is involved in the selection of the correct initiator tRNA and of the correct initiation codon. IF1 can also bind 70S ribosomes and, together with IF3, may participate in the recycling of 70S ribosomes generated upon peptidyl-tRNA drop-off.

Initiation Factor 3

Translation initiation factor 3 (IF3) is a small basic protein (180 amino acids in E. coli) encoded by the infC gene (218). IF3 is essential to cell viability; accordingly, even a slight reduction in the expression of the factor causes a measurable decrease in E. coli growth rate (186, 216).

Structure of IF3

IF3 can be cleaved into two halves, each of which is folded as a stable domain (60, 120). The two halves are connected by a lysine-rich hydrophilic linker. The positively charged nature and the length (about 15 residues) of this linker peptide are strongly conserved in bacteria. Neutron-scattering and NMR studies showed that the linker peptide accounts for separation of the two domains by about 45 Å (120, 165). The structures of both the N- and C-terminal domains of IF3 have been solved by either X-ray crystallography (B. stearothermophilus IF3 [14]) or NMR (E. coli IF3 [63, 64]). The IF3-N terminus displays a globular α/β-domain consisting of a single α-helix packed against a mixed four-stranded β-sheet. The IF3-C terminus also consists of an α/β-domain formed by two parallel helices packed against a mixed four-stranded β-sheet. The topology of the mixed four-stranded β-sheet of IF3-C, however, is different from that of IF3-N (Fig. 5). In the crystallographic structure, the linker peptide between the N and C domains of B. stearothermophilus IF3 is helical (14), whereas NMR studies have suggested that the interdomain region in E. coli IF3 is totally disordered in solution (165). It was proposed that the two structures obtained with E. coli and B. stearothermophilus IF3 are different when examined at room temperature, but they become similar if compared at their respective physiological temperatures (101, 102).

In intact IF3, both domains interact with the ribosome, albeit with different affinities (237). All known functions of native IF3 can be achieved in vitro by the isolated IF3-C domain, whereas the IF3-N domain contributes to the thermodynamic stability of the IF3–30S interaction (126, 195). In E. coli IF3, amino acids Y107 and K110, as well as a set of arginine residues of the C domain, are involved in the binding of the factor to the ribosome (39, 194).

Mapping of the binding site of IF3 to the ribosome

An early immunoelectron microscopy study localized the binding site of IF3 onto the platform of the 30S subunit (249). Recently, two structures of IF3 bound to the 30S subunit were determined. T. thermophilus 30S subunits bound to IF3 were studied by cryo-EM. A difference map calculated at 27 Å resolution from the free and IF3-bound subunits revealed three lobes of positive electron density. Two lobes were attributed to the N- and C-terminal domains of IF3. The third lobe was proposed to result from a conformational change driven by the binding of IF3. From this, the C-terminal domain of IF3 was positioned at the interface side of the platform, whereas the N-terminal domain was placed closer to the neck (151). IF3-C binding at this position may indeed impair subunit association. Moreover, a proximity of IF3-C to the anticodon stem and loop of initiator tRNA would explain the participation of this domain of the factor in initiator tRNA selection.

Another structure results from soaking crystals of the T. thermophilus 30S subunit in IF3-C. This study revealed a binding site of IF3-C at the upper end of the platform, on the solvent side of 30S (197). It must be noted, however, that this position of IF3-C may have been influenced by crystal contacts occurring at the interface surface of the platform (224, 272).

In 2001, a study using hydroxyl radical footprinting and directed probing gave further insight into a model for the interaction of IF3 with 30S (37). The model indicates proximity of IF3-C to helices 23 (656–700, T. thermophilus numbering), 24 (749–796, T. thermophilus numbering), and 45 (1484–1505, T. thermophilus numbering) of the 16S rRNA at the interface side of the platform. Such a location is supported by several biochemical studies, which suggest interaction of IF3 with bases from helices 23 and 24 (162, 170, 257) and with bases belonging to helix 45 (55, 58, 238). The model also agrees with the position of IF3-C as proposed by cryo-EM (151). IF3-C location coincides with the position of helix 69 of 23S rRNA, explaining the ability of IF3 to block subunit association. Placement of IF3-N is less clear. Dallas and Noller placed it at the E site. Such a positioning fits in with the location of the N domain at an unassigned region of electron density contiguous with the platform (third lobe; see above and reference 151). Therefore, the N and C domains of IF3 would lie on opposite faces of the initiator tRNA (37). However, IF3 would remain too distant of the tRNA molecule to envisage a direct role of the factor in tRNA discrimination. Rather, it is proposed that the binding of IF3 to the 30S subunit contributes to move bases G1338 and A1339 into intimate contact with the minor groove of the anticodon stem of tRNA_f^Met, where they may perform a steric check of the tRNA identity (37). A time-resolved chemical probing study led to the same conclusion for the position of IF3-C. However, it moves IF3-N closer to the P site (56).

Finally, tentative placement of IF3 according to the model proposed in reference 37 was attempted in the electron density of the cryo-EM structure of an E. coli 70S particle bound to mRNA, fmet-tRNA_f^Met, IF1, IF2-GDPNP, and IF3 (3). To explain the binding of IF3 to a 70S particle, a weakly associated state of the ribosomal subunits has to be imagined (56, 72, 91).

Homologies between IF3 and eIF1

The structures of human and yeast eIF1 were solved by NMR (59, 205). As shown in Fig. 5, eIF1 forms a tightly folded domain with two helices on one side of a five-stranded mixed β-sheet. This structure is reminiscent of that of IF3-C. In particular, the topology β2α2β3β4 is common to the two proteins (rmsd = 3.4 Å for 44 C-α atoms compared). The resemblance between IF3-C and e/aIF1 is less obvious than either that between IF1 and e/aIF1A, on one hand, or that between IF2 and e/aIF5B, on the other hand (121). Nevertheless, a study using directed hydroxyl radical probing unexpectedly showed that the position of eIF1 on the 40S subunit is similar to that of IF3 on the 30S subunit (139). Using this indication, eIF1 was tentatively positioned at the platform interface side in the structure of yeast 40S preinitiation complex (188). These observations, combined with the demonstration of a role of eIF1 in initiation codon selection (140, 191), suggest the universality of IF3-C/e/aIF1 in the three domains of life. Such universality was previously established for the two other bacterial initiation factors, IF1/e/aIF1A and IF2/e/aIF5B (121).

The initiator tRNA at the P site

In the 30S P site, binding of the initiator tRNA is stabilized by interactions involving the stem and loop of the anticodon (118, 179, 234, 272, 281). C1400 of 16S rRNA is stacked on base C34 of the initiator tRNA, as proposed earlier from cross-linking studies using tRNA^Val bound at the P site (201). In addition, G966 of rRNA is stacked on the ribose moiety of C34 of the tRNA. The importance of A1338 for the binding of tRNA at the P site was first suggested from cross-linking experiments (45). Indeed, A1338 and G1339 of 16S RNA ensure type I and type II minor interactions (an abundant structural motif that stabilizes RNA tertiary structures through insertion of a base in the minor groove of neighboring helices [174]) with GC base pairs 30–40 and 29–41 in the initiator tRNA. This may explain why these two G-C pairs are important for initiator tRNA binding to the P site (236). On the other side, A790 contacts the O2′ group of the ribose of A38 through interaction with its phosphate group and the O2′ group of the ribose of A38 (Fig. 6). The C-terminal tails of two proteins from the 30S subunit extend into the P site. K127 of S9 interacts with the phosphate groups of U33 and C34, while the carbonyl group of A118 of S13 contacts the phosphate group of G29. Finally, it should be remembered that the three G-C base pairs in the anticodon stem of initiator tRNA may cause a peculiar conformation of the anticodon loop (236, 274). In keeping with this idea, the crystal structures of tRNA_f^Met, free (10) or complexed with the formylase (228), showed that base A37 flips out outside the anticodon loop. On the other hand, when the tRNA is at the P site, base paired with an AUG initiation codon, the anticodon loop adopts a canonical stacked conformation (234). Possibly, the conformation of the anticodon changes upon its binding to the ribosome. Such a change might amplify the effect of a correct codon-anticodon pairing on the stability of the initiation complex, thereby enhancing the accuracy of start codon selection (see “Function and mechanism of action of IF3,” below).

Fig. 6(A) Cartoon representation of IF1 bound to the 30S subunit of T. thermophilus. The IF1 barrel is colored, with β-strands in yellow, turns in green, and helices in red. Its topology is indicated. The G530 loop is cyan, S12 is dark blue, and H44 is orange. Bases A1492 and A1493 are shown as orange sticks. The view is drawn from PDB file 1HR0 (24). (B) Cartoon representation of the anticodon stem-loop of tRNA^Phe bound to the A site of the 30S subunit of T. thermophilus. Color coding is the same as that in panel A for H44, S12, and G530 loop. mRNA is green, and ASL tRNA^Phe is yellow. The A-site codon is labeled in green, and the Phe codon is labeled in yellow. The view is drawn from PDB file 2J00 (234). (C) Interaction of the anticodon loop of initiator tRNA in the P site with mRNA and with the 30S subunit. tRNA_f^Met is shown as cyan sticks. mRNA is shown as yellow sticks. The anticodon sequence is labeled in cyan, and the AUG codon is labeled in yellow. Bases G966, C1400, A790, G1338, and A1339 are shown as sticks. S9 tail and the side chain of K127 are schematized in red. The view is drawn from PDB file 2J00 (234). Figure 6 was drawn with PyMol (DeLano Scientific LLC).

Function and mechanism of action of IF3

The involvement of IF3 in facilitating both 70S initiation complex formation and ribosome dissociation was recognized early  (51, 104, 178, 206, 213, 252). A ribosome-dissociating activity of IF3 could be observed in the presence of IF1 (189; see also “Effect of IF1 on the dissociation of 70S ribosomes,” above). However, the activity of IF3 is rather antiassociative, since the main effect of the factor is to slow down the ribosomal subunit association rate (69). Another important effect of IF3, noticed early, is to enhance the reversibility of aminoacyl-tRNA binding to mRNA-programmed 30S subunits (75, 199). In other words, IF3 accelerates both the association rate and the dissociation rate governing formation of the complexes in the P site. It was then noted that complexes containing initiator tRNA plus an AUG codon were more resistant to IF3-induced destabilization than other complexes, in such a way that IF3 eventually exerts a positive influence on the accuracy of the initiation process (75, 210). The resistance to destabilization was proposed to be related to structural features of the tRNA, in particular to the three G-C base pairs in the acceptor stem but not to the methionyl or formyl moieties of the mature initiator tRNA (87, 88, 210). Since these observations were made, numerous studies have analyzed the role of IF3 in the specificity of translation initiation (e.g., 123, 143, 156, 181, 215, 254, 258). On the whole, these studies sustain the idea that IF3 modulates the stability of one or several of the successive ribosomal complexes involved in translation initiation. More precisely, IF3 would dissociate a complex if any of the important interactions governing the stability of this complex were not filled in. Such contacts include the SD-aSD interaction, the specific binding of initiator tRNA at the P site, and the correct pairing between the tRNA and the start codon on the mRNA. Notably, such a proofreading property of IF3 was turned to good account for translational autocontrol of infC, the gene of IF3 (20, 70, 123). Indeed, infC has an unusual AUU start codon (218) that enables IF3 to selectively inhibit its own initiation of translation when present at a sufficient concentration in E. coli.

Recently, two series of kinetic data further analyzed the mode of action of IF3 (6, 7, 71, 72). The first series of experiments (6, 7) concluded that the main effect of IF3 is to increase both the dissociation rate constant of aminoacyl-tRNA binding to mRNA-programmed 30S subunits and, to a lesser extent, the corresponding association rate constant. As a result, the binding of any aminoacyl-tRNA to the 30S subunit is partly destabilized. With the initiator tRNA, the obtained destabilization counterbalances the stabilizing effect of IF2. Consequently, in the presence of the three factors, the affinity of fMet-tRNA_f^Met for the 30S subunit remains comparable to that in the absence of the factors, but the dynamics of the equilibrium are much more rapid. However, a better binding affinity would only be accounted for by an effect of IF2, depending on the formyl group. In contrast, the effect of IF3 on aminoacyl-tRNA binding would occur independently of the nature of the tRNA, in such a way that initiator tRNA binding and elongator tRNA binding would be identically destabilized. A second effect of IF3 (again amplified by IF1) is to slow down the rate of 50S subunit association. In this case, however, the slowing down is less pronounced in the presence of the initiator tRNA than in the presence of an elongator one. The results also suggest that 50S subunit docking requires prior release of IF3 from the 30S subunit, in such a way that IF3 can be considered to compete with the 50S for 30S binding. From the whole data, it was concluded that the favorable effect of IF3 on the accuracy of initiation is due to a maximization of the advantages given to initiator tRNA over the elongator ones. These advantages, obtained with the help of IF2 and IF1, concern both tRNA binding to the 30S and subunit docking (6). Such advantages depend on both the formyl group and some other features of the initiator tRNA. IF3 would not directly recognize these features but would exert its accuracy-enhancing effect by uniformly destabilizing the binding of all tRNAs to the 30S subunit and by decreasing the rate of ribosomal subunit docking (6). Indeed, the net result of this is to kinetically maximize tRNA selectivity by avoiding wrong “out of equilibrium” 30S initiation complexes (30SIC) to be erroneously stabilized upon the joining of a 50S subunit. Other observations are in keeping with such a mode of action of IF3. First, a single mutation in the IF3 N-terminal domain was shown to affect translation initiation at several levels (143). Second, it was observed that the presence of IF3 slowed down by a factor of 5 the rate of association of 50S subunits with the 30SIC (160).

A second series of experiments (71, 72) was performed with a different mRNA and under different experimental conditions in comparison with the first series. It demonstrated minor effects of the presence or absence of IF3 on the rate of 30SIC formation using either an AUG or an AUU initiation codon. However, with the incorrect AUU initiation codon, IF3 dramatically slowed down the rate of 70SIC formation from the 30SIC. More precisely, the step selectively affected by IF3 was the conversion of a labile 70S complex into a more stable 70SIC. The whole data led to the proposal that, on the labile 70S complex, IF3 remains bound in an intermediary position but must be released to achieve stabilization of the 70SIC through formation of new bridges. With improper codon-anticodon pairing, the release of IF3 and consequently the stabilization of the 70SIC would be much slower (72). This model is compatible with either a direct action of IF3 on initiator tRNA recognition or an indirect one. Notably, this model implies that IF3 remains bound to the ribosome after docking of the 50S subunit, while the results described in the preceding paragraph (6) favor the idea that IF3 leaves before ribosomal subunit joining.

Although direct inspection by IF3 of features of the initiator tRNA and of its pairing remains possible, many recent studies favor the idea that the factor exerts an indirect effect mediated by the ribosome itself. As described above (see "The initiator tRNA at the P site,” above), in addition to 16S rRNA, two ribosomal proteins, S9 and S13, interact with the P-site tRNA. Deletion of the C-terminal tail of S13 caused an unspecific decrease of tRNA affinity for the P site (97). In contrast, deletion of the C-terminal tail of S9 selectively affected the binding of those tRNAs whose anticodon stem sequences are the most divergent from that of initiator tRNA (97). More precisely, the mutated 30S subunits bound tRNAs whose anticodon stem sequences contained at least two of the three G-C pairs of initiator tRNA-displayed affinities comparable to those of wild-type subunits, while binding of other tRNAs was greatly diminished. This strongly indicates that S9 masks the natural selectivity of the P site for the initiator tRNA. Taken together with the kinetic results (6), these observations led to the proposal that S9 might decrease the association and dissociation rate constants governing the interaction of peptidyl-tRNAs at the P site and stabilize their binding during elongation of translation. However, during the initiation process, the action of S9 must be relieved to improve selectivity of the P site toward initiator tRNA. This selectivity is provided by the interactions of G1338 and A1339 with the G-C pairs of the initiator tRNA (97, 127) (see “The initiator tRNA at the P site,” above). An attractive candidate to neutralize the action of S9 obviously is IF3 (6).

On the other hand, in the presence of the 50S subunit, the selectivity observed for the S9-tail-deleted mutant vanishes. In line with this result, the binding of tRNA on a 30S subunit carrying a double substitution G1338C:A1339C can be rescued in vitro upon addition of the 50S subunit (127). These observations are in agreement with the fact that initiation of translation takes place on the small ribosomal subunit. The 30S subunit carries an inherent selectivity toward initiator-like tRNAs, even in the absence of the initiation factors. As proposed by Lancaster and Noller (127), such an RNA-based selectivity may have been present in primitive ribosomes. Since this selectivity is masked in 70S particles, dissociation of ribosomes into subunits may have been the sole function of early initiation factors (127).

Section summary

IF3 performs two distinct roles. First, it impairs association of 30S and 50S subunits and may also dissociate 70S subunits in the presence of IF1. Second, IF3 is closely involved in the selection of the correct start codon. Despite a large number of studies, the mode of action of IF3 is not fully understood and is still a subject of controversy. Selection by IF3 occurs thanks to the presence of three G-C base pairs characteristic of the anticodon stem of the initiator tRNA. These base pairs are recognized by G1338 and A1339 in the P site of the 16S rRNA, and the 30S subunit has therefore a natural selectivity for initiator tRNA at the P site. Ribosomal protein S9 appears to mask this selectivity, and IF3 may act by neutralizing the action of S9.

SNAPSHOTS OF THE TRANSLATION INITIATION PROCESS

Four cryo-EM structures of translation initiation complexes are available to date (3, 171, 242). The structure of a 30S initiation complex was determined in the presence of IF1, IF2-GTP, the initiator tRNA, and mRNA (242). In this structure, domains I (G domain) and II of IF2 interact with helices H5 and H14 of the 16S RNA, while its domain IV interacts with the acceptor end of the P-site-bound tRNA. Compared with the structures of 70S ribosome:tRNA complexes (234, 281, 282), the 30S structure shows that initiator tRNA is distorted, with the anticodon loop bent toward the mRNA. Moreover, the acceptor stem is shifted toward the E site and the CCA extremity is lifted ~15 Å away from the peptidyltransferase center. This tRNA state was termed the 30S P/I state (242). A small lobe of electron density close to the decoding A site is attributable to IF1. This corresponds to a placement of IF1 identical to that observed in the crystallographic structure (24). Considering these positions, IF2 is too far away from IF1 to envisage a direct interaction between the two factors.

The structure of an E. coli 70S initiation complex was obtained in the presence of mRNA, fMet-tRNA_f^Met, IF1, IF2-GDPNP (a nonhydrolyzable GTP analogue) and IF3 (3). IF2 interacts with both ribosomal subunits, in keeping with its subunit-joining function. Moreover, the nucleotide-binding domain I of IF2 contacts the 50S subunit in the vicinity of the GTPase-associated center, in a manner similar to EF-Tu and EF-G. Domain IV of IF2 is in contact with the initiator tRNA, but the interaction precludes classical P-site binding of the initiator tRNA. This led to the definition of a 70S P/I site. The 30S P/I state later observed in the 30S initiation complex (242) is different but reminiscent of this 70S P/I site. The position of IF1 occluding the A site, as observed in the crystallographic structure (24), is very close to domain II of IF2.

Two structures describe a T. thermophilus 70S particle with mRNA, fMet-tRNA_f^Met, and IF2 bound to either GMPPCP (another GTP analogue) or GDP (171). In both structures, determined in the absence of IF1, the tRNA is found in the classical P site (234, 281, 282) and shows no contact with IF2. Because of such a release of the tRNA from IF2, the structure in the presence of GDPCP might be representative of the GDP plus P_i state of IF2 (171), occurring after GTP hydrolysis and before IF2 release from the initiation complex (71, 262). Notably, the placement of IF1, as deduced from the crystallographic data (24), is not possible without clashes with IF2. Therefore, this structure might represent a post-IF1-release state. In the presence of GDP, IF2 steps further from the tRNA molecule and shows less contact with the ribosome, possibly in a “ready-to-leave” conformation (171).

These four cryo-EM structures likely represent four snapshots of the whole initiation process. They show different placements of IF2. According to the various positions of IF2, direct interaction of this factor with IF1 may be envisaged or not. Clearly, more instant views of the initiation process are necessary to fully understand the dynamics of all the events and to possibly highlight interactions between IF1 and IF2. Such interactions may be transient, occurring only at certain stages.

TRANSLATION OF LEADERLESS mRNAS

Leaderless mRNAs (lmRNAs) are quite rare in E. coli. In lmRNAs, the AUG start codon is located at or very close to the 5′ end. Such mRNAs appear to require no initiation signal other than the initiation codon itself (18), although translation may be facilitated by the lack of secondary structures downstream from the AUG (148). Several pieces of evidence, obtained both in vitro and in vivo, strongly indicate that initiation of translation of lmRNAs can be done with 70S ribosomes (e.g., references 164, 183, and 264). Furthermore, IF2 selectively stimulates translation of lmRNAs (73), while IF3 has the opposite effect (74, 258). Finally, translation of lmRNAs can proceed in vitro with 70S ribosomes and fMet-tRNA_f^Met, even in the absence of initiation factors, provided that excess initiator tRNA is present (264).

The available data appear consistent with an initiation process involving recognition of lmRNAs by 70S-IF2-initiator tRNA complexes, followed by pairing of the tRNA anticodon with the AUG codon (107). In this case, the effect of IF2 would be due to its stimulation of both tRNA recruitment and 50S subunit joining, while IF3 may act by facilitating the dissociation of ribosomal subunits. The critical role of the codon-anticodon pairing is underlined by the rather strict dependence on an AUG start codon, compared with GUG or UUG (182). However, a 5′-terminal AUG codon may also be directly recognized by 70S ribosomes, independently of the initiator tRNA (18).

Recently, it has been observed that 61S ribosomal particles, lacking at least 6 ribosomal proteins including S1 and S21, were induced upon treatment of E. coli  with the antibiotic kasugamycin (107). These 61S ribosomes selectively translate lmRNAs. Interestingly, S2 and S21 are present in the polysomes translating lmRNAs in the presence of kasugamycin. Indirect pieces of evidence have suggested that this association would reflect direct binding of the ribosomal proteins to lmRNAs (107).

Despite a great deal of evidence in favor of an initiation of lmRNA translation by 70S ribosomes in E. coli, a second pathway involving a 30S-IF2-initiator tRNA complex has not yet been completely ruled out (163). Further investigation is therefore required to establish which pathway is indeed used in vivo. This open question is reminiscent of another question concerning possible translation reinitiation by 70S ribosomes at distal cistrons in polycistronic mRNAs (245). Finally, lmRNAs appear to be much more frequent in some archaea, such as Sulfolobus solfataricus, than they are in E. coli (261). Such lmRNAs were shown to bind 30S ribosomal subunits in the presence of initiator tRNA (13).

CONCLUSION AND PERSPECTIVES

A wide range of strategies, including cross-linking studies, use of mutated ribosomes in vivo and in vitro, reconstitution of translation systems in vitro, and kinetic and thermodynamic studies, have led to a fairly precise view of the whole translation initiation process. Moreover, a view of translation initiation at the structural level is beginning to emerge. More snapshots are required to fully identify the sequence of events along this multistep process and to understand the exact role of the initiation factors at the atomic level.