Mechanism and Regulation of Protein Synthesis in Saccharomyces cerevisiae

mRNA features in translation initiation

In addition to translation factors, mRNA features also contribute to formation of a translating 80S ribosome. While the most important feature of an mRNA is the open reading frame (ORF), other parts of the mRNA have significant impacts on protein synthesis. Nearly all yeast proteins are initiated with methionine encoded by an AUG codon. In addition, in almost all cases, protein synthesis starts at the first AUG codon from the 5′ end of the mRNA. To date, only a few exceptions to these rules have been identified, and interestingly several of the exceptional mRNAs are subject to translational regulation or encode proteins that are targeted to more than one subcellular compartment (Hinnebusch 2011).

Initiator methionyl-tRNA

The tRNA_i^Met performs a unique role in protein synthesis. Distinct initiator and elongator tRNAs are used to incorporate methionine at the start codon vs. internal AUG codons in ORFs, respectively. Yeast contain four to five IMT genes encoding tRNA_i^Met and five EMT genes encoding elongator methionyl-tRNA (tRNA_e^Met) (Astrom et al. 1993). Whereas both sets of tRNAs contain a 5′-CAU-3′ anticodon, nucleotide and post-transcriptional modification differences restrict the function of the tRNAs to initiation vs. elongation. Swapping nucleotides between tRNA_i^Met and tRNA_e^Met has provided insights into the critical determinants for tRNA^Met function in initiation vs. elongation. Functionally important features of tRNA_i^Met include: (1) A1:U72 and C3:G70 base pairs in the acceptor stem; (2) A54 and A60 in the T loop; and (3) three G:C base pairs in the anticodon stem (positions 29–31:39–41). In addition, the nucleotide A54 and an O-ribosyl phosphate modification of A64 restrict tRNA_i^Met from functioning in translation elongation (Figure 2) (von Pawel-Rammingen et al. 1992; Astrom et al. 1993; Astrom and Bystrom 1994).

Substitution of the A1:U72 base pair in tRNA_i^Met by G1:C72, as found in tRNA_e^Met, impaired yeast cell growth (von Pawel-Rammingen et al. 1992; Astrom et al. 1993), binding of Met-tRNA_i^Met to eIF2 (Farruggio et al. 1996; Kapp et al. 2006), and TC binding to the 40S ribosome (Kapp et al. 2006). Thus the identity of this base pair contributes both to TC formation and to later steps in the initiation pathway. Three consecutive G:C base pairs in the anticodon stem are important for tRNA_i^Met function in bacteria (Varshney et al. 1993; Mandal et al. 1996), and their critical role in eukaryotes has only recently been revealed (Dong et al. 2014). Disrupting the G31:C39 base pair in the anticodon loop altered the accuracy of translation start site selection in a manner that was sensitive to the presence of the A54 residue in the T loop (Dong et al. 2014). As the C3:G70 base pair in the acceptor stem likewise contributed to the accuracy of translation start site selection (Dong et al. 2014), and all of these mutations affected the binding of the eIF2–GTP–Met-tRNA_i^Met ternary complex to the 40S ribosome, albeit in distinct ways, these results indicate that conserved nucleotides of the tRNA_i^Met contribute to the accuracy of translation start site selection.

An additional important determinant in tRNA_i^Met is a 2-O-ribosyl phosphate modification of A64. In yeast strains lacking Rit1, the enzyme that catalyzes the modification, tRNA_i^Met can function in translation elongation (Astrom and Bystrom 1994). Interestingly, domain III of EF-Tu, and by analogy of the eukaryotic elongation factor eEF1A, contacts the T loop of the bound tRNA (Nissen et al. 1995). The O-ribosyl phosphate modification of position 64 in the T loop would be expected to sterically interfere with Met-tRNA_i^Met complex formation with eEF1A. Hence, it is thought that this modification restricts tRNA_i^Met function to initiation and thus prevents competition for methionyl-tRNA between translation initiation and elongation. Consistent with this idea, deletion of RIT1 exacerbated the growth defect in strains with mutations in eIF2 or tRNA_i^Met; and this growth defect was partially rescued by overexpression of tRNA_i^Met and further exacerbated by overexpression of eEF1A (Astrom et al. 1999).

While many nucleotides in tRNAs are post-transcriptionally modified (for example by methylation, conversion to pseudouridine, etc.), it is noteworthy that tRNA_i^Met appears to be especially sensitive to these modifications. Most of the 11 modifications of yeast tRNA_i^Met are nonessential; however, loss of the m¹A58 modification destabilizes tRNA_i^Met and impairs translation initiation (Anderson et al. 1998, 2000). The Gcd10/Gcd14 complex catalyzes methylation of A58, and inactivation of GCD10 or GCD14 results in turnover of tRNA_i^Met by the Trf4/Rrp6 pathway (Kadaba et al. 2004).

Ternary complex formation

The translation factor eIF2 is responsible for binding Met-tRNA_i^Met to the 40S ribosome. A TC is formed between Met-tRNA_i^Met and the GTP-bound form of eIF2. The eIF2 is a heterotrimeric complex consisting of α (Sui2), β (Sui3), and γ (Gcd11) subunits. The yeast eIF2α (SUI2) and eIF2β (SUI3) genes were first discovered by Donahue et al. (1988) in a screen for mutations that suppress the histidine auxotrophy of his4-303 strains in which an AUU codon is substituted for the initiating AUG codon of the HIS4 gene. Spontaneous mutations in unlinked sui (suppressors of initiator codon) genes, including SUI2 (Cigan et al. 1989) and SUI3 (Donahue et al. 1988), enable translation to initiate at an in-frame UUG codon that normally encodes Leu as the third residue in HIS4 (Donahue et al. 1988). As discussed below, analysis of Sui⁻ mutations in eIF2 and other translation factors has provided insights into the mechanism of start codon selection during translation initiation.

Structures of yeast (Dhaliwal and Hoffman 2003; Hussain et al. 2014; Llacer et al. 2015), archaeal (Schmitt et al. 2012), and mammalian eIF2α (Ito et al. 2004) revealed that the protein consists of three domains: an N-terminal OB-fold domain and a central α-helical domain that are connected through a flexible linker to a C-terminal α/β domain that binds to eIF2γ (Figure 3A) (Schmitt et al. 2012; Hussain et al. 2014; Llacer et al. 2015). A key mode of translational control in yeast and other eukaryotes involves phosphorylation of eIF2α. The yeast kinase Gcn2, which is conserved in all eukaryotes, phosphorylates the conserved Ser51 residue in a mobile loop of the OB-fold domain (Dever et al. 1992).

Whereas translation of the GCN4 mRNA in yeast is typically repressed by the presence of uORFs in the mRNA leader, phosphorylation of eIF2α enables ribosomes to bypass the inhibitory uORFs and initiate translation at the GCN4 ORF (reviewed in Hinnebusch 2005). Yeast lacking GCN2 are unable to grow under amino acid starvation conditions due to the failure to derepress GCN4 expression (Wek et al. 1989, 1990). Gcd⁻ mutations, including mutations that impair eIF2 function (Williams et al. 1989), derepress GCN4 expression in the absence of GCN2, mimicking the effect of eIF2 phosphorylation in reducing TC assembly (Hinnebusch 2005).

The C-terminal half of yeast eIF2β shows significant sequence homology to archaeal aIF2β and consists of three elements: an N-terminal α-helix, a central helix–turn–helix domain, and a C-terminal zinc-binding domain (Figure 3A) (reviewed in Schmitt et al. 2010). The N-terminal α-helix, which is unstructured in the free form of aIF2β, binds to the backside of the aIF2γ GTP-binding (G) domain in the aIF2 complex (Sokabe et al. 2006; Yatime et al. 2007). Point mutations in this helix of yeast eIF2β, as well as in the docking site on yeast eIF2γ, disrupt eIF2 complex formation and confer Gcd⁻ and Sui⁻ phenotypes (Hashimoto et al. 2002; Borck et al. 2012). In the archaeal aIF2 complex, the C-terminal zinc-binding domain of aIF2β packs against the central α–β domain (Sokabe et al. 2006; Yatime et al. 2007). While the function of the zinc-binding domain has not been resolved, removal of this domain impairs RNA binding to isolated yeast eIF2β (Laurino et al. 1999) and confers a dominant Gcd⁻ and recessive lethal phenotype (Castilho-Valavicius et al. 1992), while point mutations confer a dominant Sui⁻ phenotype (Donahue et al. 1988; Castilho-Valavicius et al. 1992).

The N-terminal half of eIF2β is not present in the archaeal protein. Key features of this portion of eIF2β are three elements referred to as K-boxes K1, K2, and K3, each containing seven Lys residues and one Ser or Thr residue. Whereas deletion of any single or two K-boxes does not affect cell viability, removal of all three K-boxes is lethal (Asano et al. 1999; Laurino et al. 1999). Consistent with these findings, substituting Ala residues in place of the K3 Lys residues in a SUI3 allele lacking K1 and K2 was also lethal. In contrast, substituting Arg residues in place of the K3 Lys residues in the same allele was viable and had no impact on cell growth (Laurino et al. 1999). Thus, the positively charged character of at least one K-box is required for cell viability. Biochemical analyses revealed that removal of the K-boxes impairs mRNA, but not Met-tRNA_i^Met, binding to isolated eIF2 complexes (Laurino et al. 1999). Moreover, mutating the K-boxes in eIF2β impairs the binding of isolated eIF2β, as well as the eIF2 complex, with both the eIF2 GTPase stimulatory factor eIF5 and the catalytic ε-subunit of the eIF2 guanine-nucleotide exchange factor (GEF) eIF2B (Asano et al. 1999). A bipartite element consisting of acidic and aromatic amino acids is conserved at the C termini of eIF5 and eIF2Bε and mediates the K-box-dependent interaction with the N terminus of eIF2β (Asano et al. 1999).

The γ-subunit of eIF2, encoded by GCD11, was first identified based on the Gcd⁻ phenotype of several mutants (Hannig et al. 1993). Interestingly, mutations in eIF2γ were independently isolated in a screen for Sui⁻ mutants (Huang et al. 1997). Consistent with these findings, the GCD11-R510H mutant, originally isolated based on its ability to derepress GCN4 expression, also confers a Sui⁻ phenotype (Dorris et al. 1995). The eIF2γ protein consists of three domains: an N-terminal GTP binding domain and β-barrel domains II and III (Figure 3A). Based on structural studies of the archaeal and yeast complexes, eIF2γ is the keystone of the eIF2 complex with separate docking sites for the eIF2α and eIF2β subunits (Schmitt et al. 2012; Hussain et al. 2014; Llacer et al. 2015). The incorporation of eIF2γ in the eIF2 complex is dependent on the apparently eIF2-specific chaperone Cdc123 (Perzlmaier et al. 2013). The amino acid sequence and structure of eIF2γ and aIF2γ show striking similarity to elongation factor EF-Tu from bacteria (Hannig et al. 1993; Schmitt et al. 2002; Roll-Mecak et al. 2004). Whereas EF-Tu binds diverse aminoacyl-tRNAs (aa-tRNAs) to the ribosomal A site, eIF2 specifically binds Met-tRNA_i^Met to the ribosomal P site. The structure of Phe-tRNA bound to EF-Tu revealed that the amino acid and acceptor stem of the tRNA bind in a pocket formed between the G domain and domain II. Supporting the notion that eIF2γ uses a similar pocket for Met-tRNA_i^Met binding, the slow-growth phenotype of the gcd11-Y142H mutant, which alters a residue in the proposed Met-tRNA_i^Met binding pocket, was partially suppressed by overexpression of tRNA_i^Met, and purified eIF2 complexes containing the mutant eIF2γ subunit showed defects in Met-tRNA_i^Met binding (Dorris et al. 1995; Erickson and Hannig 1996; Shin et al. 2011). Thus, at least the acceptor stem and amino acid binding site appears to be shared between eIF2 and EF-Tu. In contrast, a contact between the body of the tRNA, especially the T stem, and domain III of EF-Tu is apparently not conserved in eIF2γ (Nissen et al. 1995; Sanderson and Uhlenbeck 2007a,b; Shin et al. 2011; Schmitt et al. 2012; Hussain et al. 2014; Llacer et al. 2015). Instead, hydroxyl radical probing experiments and cryo-EM structures of 48S PICs indicated that domain III of eIF2γ projects toward, but does not contact, helix h44 on the subunit interface surface of the 40S ribosomal subunit (Shin et al. 2011).

Purified yeast eIF2 binds either GTP (K_d ∼1.7 µM) or GDP (K_d ∼0.02 µM) (Kapp and Lorsch 2004a). As for a number of G proteins, this 100-fold higher affinity for GDP relative to GTP introduces the requirement for eIF2B to recycle eIF2–GDP complexes to the functional eIF2–GTP form. Whereas eIF2–GTP complexes bind Met-tRNA_i^Met (K_d ∼9 nM) to form a ternary complex, eIF2–GDP binary complexes are defective for Met-tRNA_i^Met binding (K_d ∼150 nM) (Kapp and Lorsch 2004a). Thermodynamic coupling between GTP and Met-tRNA_i^Met binding to eIF2 results in a 10-fold increase in GTP binding affinity in the presence of Met-tRNA_i^Met (GTP K_d ∼0.2 nM) (Kapp and Lorsch 2004a). Consistent with this biochemical result, the slow-growth phenotype of an eIF2γ-K250R mutation, which impairs GDP and GTP binding to eIF2, is suppressed by overexpression of tRNA_i^Met (Erickson and Hannig 1996).

As noted above, conserved features of tRNA_i^Met contribute to ternary complex formation. However, the Met on Met-tRNA_i^Met appears to be the most important determinant for TC formation. Deacylation of Met-tRNA_i^Met decreases its affinity for binding to eIF2 by >10-fold (K_d ∼130 nM), comparable to the binding of Met-tRNA_i^Met to eIF2–GDP (Kapp and Lorsch 2004a). It is postulated that the thermodynamic coupling between eIF2 and the methionine residue on Met-tRNA_i^Met serves to ensure that translation initiates exclusively with Met.

Interestingly, the eIF2γ-K250R mutation in addition to weakening GDP binding enables cell survival in the absence of eIF2α (Erickson et al. 2001). The growth of the gcd11-K250R sui2Δ strain is further enhanced by overexpression of tRNA_i^Met (IMT4) and by overexpression of gcd11-K250R and SUI3 (Erickson et al. 2001). As weakening GDP binding to eIF2 enables elimination of eIF2α, these findings suggest that eIF2α plays a role in stimulating the eIF2B-catalyzed guanine nucleotide exchange on eIF2.

43S PIC formation

Binding of the eIF2 TC to the 40S subunit is facilitated by the factors eIF1 and eIF1A that bind directly to the 40S ribosome (Figure 1). The factor eIF1, encoded by SUI1, is a small (108 amino acid) protein that, based on structures of the yeast or the analogous Tetrahymena factor, binds to the platform of the 40S subunit near the P site (Figure 4, A and B) (Rabl et al. 2011; Hussain et al. 2014). The factor eIF1A, encoded by TIF11, is homologous to the bacterial factor IF1 (Battiste et al. 2000; Choi et al. 2000; Olsen et al. 2003; Hussain et al. 2014). Like IF1, eIF1A binds to the ribosomal A site and likely functions, in part, to prevent Met-tRNA_i^Met binding in the A site (Figure 4, A and B) (Carter et al. 2001; Hussain et al. 2014). Cryoelectron microscopy of the yeast 40S ribosome has revealed conformational changes accompanying the binding of eIF1 and eIF1A (Passmore et al. 2007; Hussain et al. 2014). In the absence of factors, the “latch” of the mRNA entry channel, composed of 18S rRNA helices h34 in the head and h18 in the body of the 40S subunit, is closed. Binding of eIF1 and eIF1A to the 40S subunit is accompanied by rotation of the head of the subunit (Hussain et al. 2014), perhaps providing access for the Met-tRNA_i^Met and TC and by weakening of the latch interactions to enable binding of mRNA (Passmore et al. 2007; Hussain et al. 2014). Interestingly, when only eIF1A is bound to the 40S subunit, the density corresponding to the latch is stronger than that observed in the apo-40S structure. As described below, this so-called “closed” complex in the absence of eIF1 is thought to be associated with selection of the translation start codon.

Yeast eIF1 is composed of an ∼20-residue unstructured N-terminal tail (NTT) followed by an ∼88-residue folded α/β core (Figure 4A) (Reibarkh et al. 2008). The α/β core of eIF1 resembles similar domains in eIF2β, the N terminus of eIF5 (Figure 3C), and several ribosomal proteins (Reibarkh et al. 2008). In addition to binding the ribosome and regulating TC binding, eIF1 directly contacts eIF2β, the C-terminal domain of eIF5, and eIF3c. As these contacts have been mapped to distinct regions of eIF1, it is thought that eIF1 can simultaneously bind all three factors, and consistent with this idea, eIF1 can be found in a multifactor complex (MFC) with the eIF2 TC, eIF3, and eIF5 (Asano et al. 2000).

The 153-residue yeast eIF1A consists of a central OB-fold domain that resembles the bacterial factor IF1 (see Fekete et al. 2005). The core of eIF1A is buttressed on its C-terminal side by a helical region consisting of a long α2 helix and a short 3₁₀ helix. In addition, the factor has long unstructured N(∼25-residues)- and C(∼34 residue)-terminal tails (Figure 4A). In the 43S complex, the C-terminal tail (CTT) of eIF1A crosses through the P site (Hussain et al. 2014; Zhang et al. 2015). The Met-tRNA_i^Met is thus prevented from fully engaging the P site (P_in state) and instead is thought to be in a P_out state that is more conducive to scanning (Hinnebusch 2011, 2014). As discussed below, the CTT of eIF1A also interacts with the N-terminal domain (NTD) of eIF5 and with domain IV of eIF5B in subsequent steps of the initiation pathway.

Despite interacting with distinct sites, binding of eIF1 and eIF1A to the 40S subunit is thermodynamically coupled (Maag and Lorsch 2003). Moreover, both factors are required to achieve stable binding of the eIF2 TC in vitro (Algire et al. 2002). Consistent with these findings, mutations in the eIF1 core domain or in eIF1A that weaken their binding to the 40S ribosome likewise decrease the rate of TC binding in vitro and confer Gcd⁻ phenotypes in vivo (Fekete et al. 2005; Cheung et al. 2007).

Whereas eIF1 and eIF1A are critical for TC binding to the 40S in the reconstituted yeast in vitro translation system, the factor eIF3 has been reported to stabilize TC binding by only approximately twofold (Kapp and Lorsch 2004b). Yeast eIF3 is composed of five essential subunits (a/Tif32, b/Prt1, c/Nip1, i/Tif34, and g/Tif35) and one nonessential subunit (j/Hcr1) (Figure 5) (note that the unusual nomenclature of the yeast eIF3 subunits is due to the presence of additional subunits in the mammalian factor that are not present in yeast eIF3). In addition, eIF5 (Tif5) purifies stoichiometrically with tagged forms of eIF3 from yeast (Phan et al. 1998). Extensive mapping studies of protein–protein interactions have provided insights into the structure of eIF3 and its interaction with other factors (reviewed in Valasek 2012). The eIF3b/Prt1 is thought to form the primary scaffold of the multisubunit complex. The N terminus of eIF3b/Prt1 contains an RNA recognition motif (RRM) that serves as a protein–protein interaction site for eIF3j/Hcr1 as well as for the C-terminal part of eIF3a/Tif32, which resembles eIF3j/Hcr1. The central part of eIF3b/Prt1 binds to eIF3c/Nip1, and the eIF3i/Tif34 and eIF3g/Tif35 subunits bind cooperatively to the C-terminal portion of eIF3b/Prt1. Finally, the N-terminal portion of the eIF3c/Nip1 subunit binds directly to eIF1 and to eIF5, which in turn binds the eIF2 TC (Figure 5). Thus, eIF3 and in particular the eIF3b subunit plays a central role in assembly of the 43S PIC. In contrast to eIF1, eIF1A, and the TC, which bind to the intersubunit face of the 40S subunit, cryo-EM studies revealed that the core of yeast eIF3 binds to the solvent-exposed face of the 40S with arm-like projections, including the PCI domains of eIF3a and eIF3c, that bind near the mRNA entry channel reaching around to the intersubunit face of the 40S (Erzberger et al. 2014; Aylett et al. 2015; Llacer et al. 2015). Consistent with this model of the eIF3–40S complex, yeast eIF3 subunits have been found to interact with 18S rRNA and ribosomal proteins on the solvent-exposed side of the 40S subunit (Valasek et al. 2003; Kouba et al. 2012a,b). The C terminus of eIF3a/Tif32 was shown to bind to a region of 18S rRNA encompassing helices h16–h18, and in two-hybrid assays this same portion of eIF3a/Tif32 bound to ribosomal proteins Rps2 and Rps3. These interactions place eIF3a near the mRNA entry channel of the 40S subunit.

Inactivation of a temperature-sensitive eIF3b/Prt1 mutant (Phan et al. 1998; Nielsen et al. 2004) or depletion of eIF3c/Nip1 (Phan et al. 1998; Valasek et al. 2004) impairs general translation in vivo and in vitro. Moreover, extracts from these strains exhibit a defect in binding Met-tRNA_i^Met to 40S subunits that was rescued by adding back the eIF3 complex (Phan et al. 1998). The Met-tRNA_i^Met and mRNA binding defects in extracts from the prt1-1 strain were also rescued by addition of an eIF3abc, but not an eIF3big, partial complex (Phan et al. 2001). It is noteworthy that the factors eIF5, eIF1, and eIF3j/Hcr1 co-purified with the eIF3abc partial complex, raising the possibility that these latter factors contributed to the complementing activity. These results uncover a functional specialization within the eIF3 complex and they also support previous studies in mammalian systems, indicating that Met-tRNA_i^Met binding to the 40S subunit is a prerequisite for the ribosome to bind to an mRNA (see Hinnebusch 2000).

In addition to the sequential assembly of the 43S complex with eIF1 and eIF1A binding to the 40S subunit prior to association of the TC, an en masse assembly of the 43S complex has also been proposed. A MFC consisting of eIF1, eIF2, eIF3, and eIF5 plus Met-tRNA_i^Met has been isolated from cells, and in crude cell extracts the MFC can be separated from the 40S ribosome (Asano et al. 2000). It has been proposed that preassembly of the MFC facilitates proper binding of Met-tRNA_i^Met to the 40S subunit. Consistent with this hypothesis, the protein–protein interactions required for MFC integrity, including the binding of eIF1 to eIF3 (Singh et al. 2004; Valasek et al. 2004), eIF5 to eIF1, eIF2β, and eIF3c (Singh et al. 2004, 2005; Valasek et al. 2004; Yamamoto et al. 2005) and eIF2 to eIF3a (Valasek et al. 2002; Nielsen et al. 2004), are also important for protein synthesis in vivo, and mutations that disrupt eIF3c interaction with eIF1 or eIF5 confer Sui⁻ phenotypes (Valasek et al. 2004). While the in vivo data support the idea that MFC integrity is important for translation initiation, additional experiments are needed to define the function of the MFC. In particular, it is important to determine whether the MFC binds to the 40S en masse and serves as a more efficient means to bind Met-tRNA_i^Met to the 40S subunit. Alternatively, it has been proposed that the MFC might serve as a depot for the initiation factors that are critical for stable binding of Met-tRNA_i^Met to the 40S (Aitken and Lorsch 2012).

mRNA recruitment of the 43S PIC

The 5′ cap and 3′ poly(A) tail of mRNAs serve as binding sites for eIF4E and the poly(A) binding protein Pab1, respectively, that act synergistically to assist in recruiting additional translation initiation factors including eIF4G and the 43S PIC to near the 5′ end (Figure 1 and Figure 6A) (Tarun and Sachs 1995; Preiss and Hentze 1998). Yeast mRNA 5′ leader sequences are of variable length and can contain secondary structures that impede 43S binding and scanning to AUG initiation codons. As a consequence ATP-dependent RNA helicases such as eIF4A and Ded1 are recruited. Our understanding of the roles of factors in these key steps is outlined below.

AUG selection

Following binding of the 43S PIC near the 5′ end of the mRNA, it traverses in a 3′ direction inspecting for a start codon. Elegant experiments by Donahue and colleagues established that the anticodon of the Met-tRNA_i^Met in the 43S complex is primarily responsible for start codon selection. Mutation of the tRNA_i^Met anticodon from 5′-CAU-3′ to 5′-CCU-3′ enabled ribosomes to synthesize His4 when the HIS4 mRNA start codon was mutated from 5′-AUG-3′ to 5′-AGG-3′ (Cigan et al. 1988a). Moreover, insertion of a 5′-AGG-3′ codon in the HIS4 mRNA leader upstream and out-of-frame with the 5′-AGG-3′ codon at the HIS4 start site blocked His4 production (Cigan et al. 1988a). This latter result supports the model that the anticodon of Met-tRNA_i^Met in the scanning 43S complex inspects the mRNA in a base-by-base manner to select the translation start site.

This importance of codon–anticodon match in start codon selection was further supported by studies examining the kinetics and thermodynamics of 48S PIC formation in reconstituted yeast in vitro translation assays. Point mutations that altered the second or third positions of the AUG start codon on the mRNA dramatically lowered the affinity of Met-tRNA_i^Met binding in the 48S PIC. This binding defect was suppressed by mutations in the tRNA_i^Met anticodon that restored base-pairing interactions with the mRNA. As the start codon mutations in the mRNA mainly affected the on rate for Met-tRNA_i^Met binding, and not the off rate, it was proposed that 48S PIC formation is accompanied by a conformational change that locks in Met-tRNA_i^Met binding. Accordingly, in this closed state, Met-tRNA_i^Met is stably bound to the 40S subunit and fixed on the translation start codon of the mRNA (Kolitz et al. 2009).

In addition to the Met-tRNA_i^Met, translation factors play key roles in the transition of the 40S subunit from its open, scanning-competent state to the closed, scanning-arrested state following start codon selection (reviewed in Hinnebusch 2011, 2014). Genetic screens in yeast have provided key insights into the factors contributing to start codon selection. Spontaneous Sui⁻ mutations that enhance initiation from a UUG codon were isolated in eIF1 (Yoon and Donahue 1992), all three subunits of eIF2 (Donahue et al. 1988; Cigan et al. 1989; Huang et al. 1997), and in eIF5 (encoded by TIF5) (Huang et al. 1997). In subsequent directed screens Sui⁻ mutations have also been isolated in eIF1A (Fekete et al. 2007), eIF3 subunits (Valasek et al. 2004; Chiu et al. 2010; Elantak et al. 2010; Karaskova et al. 2012), and in 18S rRNA (Nemoto et al. 2010). In contrast to the Sui⁻ mutations, which relax the stringency for start codon selection, a second class of mutations enhances start codon selectivity. The Ssu⁻ (suppressor of Sui⁻) mutations block the ability of Sui⁻ mutations to enhance initiation at a UUG codon in a mutant HIS4 allele (Asano et al. 2001; Fekete et al. 2007; Saini et al. 2010). In general, Sui⁻ mutations are thought to block scanning and promote conversion of the 40S subunit to its closed, scanning arrested conformation (P_in). In contrast, the Ssu⁻ mutations promote scanning and stabilize the open conformation of the 40S subunit (P_out).

Biochemical analyses of the Sui⁻ and Ssu⁻ mutant forms of initiation factors have provided insights into the mechanism of start codon selection. The dominant SUI5-G31R mutation in eIF5 was reported to alter the release of Pi from GTP following its hydrolysis by eIF2 (Saini et al. 2014), and the SUI3-2 (S264Y) mutation in eIF2β and the SUI4 (GCD11-N135K) mutation in eIF2γ were reported to increase the intrinsic (eIF5 independent) GTPase activity of eIF2 (Huang et al. 1997). Thus, it was proposed that premature GTP hydrolysis by eIF2 would enable release of Met-tRNA_i^Met to the P site in the absence of perfect codon–anticodon base pairing. The eIF5-G31R and eIF2β-S264Y mutations also stabilize 48S PICs at UUG codons in the presence of nonhydrolyzable GTP, indicating these mutations stabilize the closed P_in state of the 48S PIC (Martin-Marcos et al. 2014; Saini et al. 2014). The identification of Sui⁻ mutations in the CTT and Ssu⁻ mutations in the NTT of eIF1A likewise implicate these segments in stabilizing the P_out or P_in states of Met-tRNA_i^Met binding, respectively (Fekete et al. 2007; Saini et al. 2010).

Sui⁻ mutations in eIF1 have been found to weaken eIF1 binding to the 40S subunit, consistent with their recessive phenotype and with the notion that eIF1 dissociation from the 48S complex is required for start codon selection (Cheung et al. 2007; Martin-Marcos et al. 2013). Moreover, overexpression of eIF1 suppresses Sui⁻ mutations in other translation factors, indicating that eIF1 dissociation from the 48S complex is a key commitment step in start codon selection (Valasek et al. 2004; Cheung et al. 2007; Martin-Marcos et al. 2011; Martin-Marcos et al. 2013; Martin-Marcos et al. 2014). Cryo-EM structures of 48S PICs in open and closed states revealed a clash between eIF1 and Met-tRNA_i^Met in the P_in state (Llacer et al. 2015). This clash likely underlies the role of eIF1 in blocking initiation at non-AUG codons and evokes eIF1 release on AUG codon recognition (Figure 4, A and B). In addition to these genetic and structural studies, kinetic experiments have highlighted the critical gatekeeper function of eIF1 in regulating start codon selection. Recognition of the AUG start codon by the 48S PIC induces a conformational change that accelerates eIF1 dissociation, which in turn enables release of Pi from eIF2–GDP (Figure 4B) (Algire et al. 2005; Maag et al. 2005). Thus, eIF1 dissociation and the attendant release of Pi, and not simply GTP hydrolysis by eIF2 (Maag et al. 2005), is the irreversible step that commits the ribosome to initiate at the selected codon.

The 405-residue factor eIF5 folds into functionally distinct N- and C-terminal domains (Figure 3B). The N-terminal domain of eIF5 resembles eIF1 and like the C-terminal domain of eIF2β possesses a Zn-finger element (Conte et al. 2006). In contrast, the α-helical C-terminal domain of eIF5 folds into a HEAT repeat with structural similarity to the HEAT domains in the C terminus of eIF2Bε and in eIF4G (Wei et al. 2010). The N-terminal domain of eIF5 directly binds the G domain of eIF2γ (Alone and Dever 2006), and mutation of Arg15 in eIF5 confers a lethal phenotype and significantly impairs the ability of eIF5 to stimulate GTP hydrolysis by eIF2 (Das et al. 2001; Algire et al. 2005), supporting the notion that eIF5 functions as a GTPase activating protein (GAP) for eIF2. In accord with the gatekeeper function of eIF1, it is proposed that eIF1 release following start codon recognition enables eIF5 to move into the vacated space and thereby stimulate Pi release from eIF2–GDP+Pi (Nanda et al. 2009). Consistent with this proposed movement of eIF5 into the space previously occupied by eIF1, and thus closer to eIF1A, the SUI5-G31R mutation in eIF5 was found to strengthen eIF1A interaction with the PIC at a UUG codon (Maag et al. 2006). Moreover, mutations in the eIF1A CTT uncouple Pi release from eIF1 dissociation (Nanda et al. 2013). Thus the eIF5 N-terminal domain appears to be intimately involved in the structural rearrangements in the scanning ribosome upon start codon selection.

The C-terminal HEAT domain in eIF5 binds to eIF1, the NTT (K-boxes) of eIF2β, and to eIF3c/Nip1 (Yamamoto et al. 2005). Mutations in the eIF5 C-terminal domain that disrupted its interaction with both eIF1 and eIF2β conferred an Ssu⁻ phenotype and destabilized the closed state of the 48S complex (Luna et al. 2012). Importantly, this mutation did not affect the ability of eIF5 to promote GTP hydrolysis by eIF2 (Luna et al. 2012), further strengthening the notion that GTP hydrolysis occurs prior to the step controlling start codon selection (Algire et al. 2005; Maag et al. 2005).

Taking into account the results from the various genetic and biochemical studies on the translation factors that participate in start codon selection, a model can be proposed wherein the factors eIF1, eIF1A, eIF3, eIF5, and the eIF2 TC are bound to the 40S subunit as it scans the mRNA (Figure 4B). In this open, scanning-competent complex the Met-tRNA_i^Met resides in the P_out state with eIF1 bound adjacent to the P site and eIF1A bound in the A site with its N- and C-terminal tails projecting into the P site. Both eIF1 and the eIF1A CTT prevent full accommodation of Met-tRNA_i^Met into the P_in state. When the scanning complex encounters a start codon, base-pairing interactions between the anticodon of Met-tRNA_i^Met and the start codon triggers entry of the Met-tRNA_i^Met to the P_in state. This movement is accompanied by displacement of eIF1 and movement of the eIF1A CTT toward eIF5. These factor movements trigger Pi release from eIF2, a critical commitment step in start codon selection (Algire et al. 2005), and conversion of the 43S PIC to its closed, scanning-arrested state (Hinnebusch 2011, 2014). In accord with this model, recent cryo-EM structures of 48S complexes in the open and closed states have revealed conformational changes in the 40S subunit as well as interaction of the eIF1A NTT with the Met-tRNA_i^Met–AUG codon duplex, and eIF2, eIF1A, and ribosomal proteins with the mRNA and start codon context nucleotides (Hussain et al. 2014; Llacer et al. 2015). Upon start codon selection, constriction of the mRNA channel and tightening of the P site are thought to block further scanning by the PIC.

Subunit joining

Following eIF1 and Pi release, the 48S PIC is in a closed conformation with Met-tRNA_i^Met fully accommodated in the P site. Accompanying these changes, eIF1A binding to the 48S complex becomes tighter. It is unclear when the eIF2–GDP complex, eIF5 and eIF3 dissociate from the PIC; however, it is clear that based on its binding site on the intersubunit face of the 40S subunit, eIF2 must dissociate prior to 60S subunit joining.

The factor eIF5B, encoded by FUN12 and an ortholog of the bacterial translation factor IF2, promotes 60S subunit joining (Choi et al. 1998; Pestova et al. 2000). eIF5B is 1002 amino acids in length and contains a GTP-binding domain near the center of the protein. Deletion of FUN12 severely impairs yeast cell growth and causes a loss of polysomes, consistent with a defect in translation initiation (Choi et al. 1998). Removal of the N-terminal ∼400 residues of eIF5B confers no growth defect in vivo and the truncated protein catalyzes subunit joining in vitro (Lee et al. 1999; Shin et al. 2002). The crystal structure of aIF5B, the archaeal ortholog of eIF5B, revealed a chalice-shaped protein with the G domain, domain II, and domain III, forming the cup of the chalice which is connected to the base, domain IV, by a long α-helix (Roll-Mecak et al. 2000; Kuhle and Ficner 2014a). Directed hydroxyl radical mapping studies of eIF5B–80S complexes placed domain II of yeast eIF5B near 18S rRNA helix h5 of the 40S subunit (Shin et al. 2009), consistent with results of a recent cryo-EM structure of eIF5B bound to an 80S initiation complex (Fernandez et al. 2013). This binding site is compatible with the eIF5B G domain binding to the GTPase activation center on the 60S subunit, similar to the binding sites of the bacterial translational GTPases IF2, EF-Tu, and EF-G on 70S ribosomes. Whereas the C-terminal domain of bacterial IF2, which corresponds to domain IV of eIF5B, directly binds the formylmethionine (fMet) on fMet-tRNA_i^Met, direct binding of eIF5B with Met-tRNA_i^Met has not been observed. However, based on the ribosomal binding site of eIF5B and the dimensions of aIF5B, domain IV of eIF5B is thought to project across the A site to interact with Met-tRNA_i^Met in the P site. This proposed contact of eIF5B with Met-tRNA_i^Met in the P site is consistent with cryo-EM structures of the initiation complex (Fernandez et al. 2013; Kuhle and Ficner 2014b) and with the instability of 48S PICs and decreased recognition of an inhibitory upstream AUG codon (“leaky scanning”) in yeast lacking eIF5B (Lee et al. 2002). It is proposed that eIF5B binding to the closed 48S complex at an AUG codon stabilizes Met-tRNA_i^Met binding following eIF2–GDP release and promotes 60S subunit joining. In the absence of eIF5B, the Met-tRNA_i^Met is not stably bound, causing some 48S complexes to dissociate from the mRNA and others to resume scanning to downstream start sites.

Domain IV of eIF5B binds to eIF1A (Choi et al. 2000) via interaction with the last five residues at the C terminus of eIF1A (Olsen et al. 2003; Acker et al. 2006; Fringer et al. 2007). Mutation of the eIF1A C terminus impairs subunit joining and full activation of eIF5B GTPase activity in vitro (Acker et al. 2006) and impairs yeast cell growth and eIF5B binding to 40S complexes in vivo. The growth and translation initiation defects of this eIF1A mutant are suppressed by overexpression of eIF5B, indicating that eIF1A helps recruit eIF5B to 40S complexes prior to subunit joining and thereby accelerates ribosomal subunit joining (Acker et al. 2009).

Both eIF1A and eIF5B are bound to the 80S ribosome following subunit joining, and GTP hydrolysis by eIF5B is required for their release (Shin et al. 2002; Fringer et al. 2007; Acker et al. 2009). Blocking eIF5B GTPase activity, either by inclusion of nonhydrolyzable GTP analogs or by mutation of the eIF5B G domain, does not impair subunit joining (Shin et al. 2002, 2007, 2009; Acker et al. 2009); however, it does impede eIF1A release from 80S ribosomes both in vivo and in vitro (Fringer et al. 2007; Acker et al. 2009). Mutations that disrupt the GTPase activity of eIF5B severely impair yeast cell growth (Shin et al. 2002, 2007, 2009). Suppressor mutations of these eIF5B mutants either restore the factor’s GTPase activity (Shin et al. 2007) or decrease the binding affinity of eIF5B for the 80S ribosome (Shin et al. 2002, 2009), consistent with GTP hydrolysis lowering eIF5B affinity for the ribosome. It is proposed that GTP hydrolysis by eIF5B alters the conformation of the 80S to promote eIF1A release (Acker et al. 2009). As the eIF5B suppressor mutants that bypass the requirement for GTP hydrolysis show enhanced levels of leaky scanning (Shin et al. 2002), GTP hydrolysis by eIF5B might serve as a checkpoint to ensure the fidelity of subunit joining (Shin et al. 2002). Following release of eIF5B and eIF1A, the ribosome is poised with Met-tRNA_i^Met in the P site and a vacant A site available to receive the first elongator aa-tRNA. It is possible that some initiation factors including eIF3 (Szamecz et al. 2008) remain associated with the ribosome through the first few steps of elongation.

Recycling eIF2–GDP to eIF2–GTP

eIF2–GDP released from the 48S PIC following start codon selection must be converted to an active GTP-bound form to promote Met-tRNA_i^Met binding and continued rounds of translation initiation (Figure 1 and Figure 7A). This is an important step as phosphorylation of eIF2α converts eIF2 into an inhibitor of its GEF eIF2B, thereby lowering TC levels. eIF2 recycling was thought to be a single reaction involving eIF2B; however, eIF5 antagonizes eIF2B and must be displaced from eIF2 prior to nucleotide exchange (Figure 1 and Figure 7A) (Jennings and Pavitt 2014).

aa-tRNA delivery by eEF1

The eEF1A–eEF1Bαγ complex in yeast, like the analogous EF-Tu–EF-Ts complex of bacteria, provides both aa-tRNA delivery (eEF1A) and GEF (eEF1Bαγ) functions. Consistent with their strong sequence and structural similarity, the canonical translation functions and kinetic mechanisms of G proteins eEF1A and EF-Tu are very similar. In addition, studies in yeast have illuminated roles for elongation factors outside translation elongation.

eEF2 and ribosomal translocation

The unique elongation factor eEF3

eIF5A promotion of peptide bond formation

The translation factor eIF5A was originally identified as an initiation factor based on its ability to promote methionyl–puromycin synthesis (Kemper et al. 1976; Schreier et al. 1977; Benne et al. 1978; Benne and Hershey 1978), an in vitro assay designed to monitor first peptide bond synthesis. However, this assay also monitors the peptidyl-transferase activity of the ribosome. Yeast eIF5A is encoded by TIF51A (HYP2) and TIF51B (ANB1). Whereas TIF51A is expressed in aerobically grown yeast, TIF51B expression is restricted to anaerobic conditions (Lowry et al. 1983; Schnier et al. 1991). It is not clear why yeast differentially express the two forms of eIF5A that differ at only 15 of their 157 residues.

eIF5A is post-translationally modified by the conversion of a lysine residue to hypusine (Wolff et al. 2007; Dever et al. 2014). An n-butylamine group is transferred from spermidine to the ε-amino group of Lys51 in eIF5A to generate deoxyhypusine. This reaction is catalyzed by deoxyhypusine synthase, encoded by the essential DYS1 gene (Kang et al. 1995; Sasaki et al. 1996; Park et al. 1998). In a second step, the enzyme deoxyhypusine hydroxylase, encoded by nonessential LIA1, hydroxylates the 2-position of the added moiety to generate hypusine (Park et al. 2006a).

Yeast eIF5A was found to interact with the translational machinery (Jao and Chen 2006; Zanelli et al. 2006; Saini et al. 2009), and depletion of eIF5A impaired protein synthesis and yeast cell growth (Kang and Hershey 1994; Saini et al. 2009; Henderson and Hershey 2011). Analysis of polyribosome profiles of yeast depleted of eIF5A or following inactivation of a temperature-sensitive eIF5A mutant revealed retention of polysomes in the absence of cycloheximide (CHX), indicating a defect in translation elongation (Gregio et al. 2009; Saini et al. 2009). However, it has also been reported that depletion of eIF5A causes a loss of polysomes, suggesting that eIF5A may have a function in translation initiation or perhaps a specialized function in translation elongation (Henderson and Hershey 2011). Consistent with the translation elongation defect observed in the eIF5A temperature-sensitive mutant, purified eIF5A stimulated tripeptide synthesis in reconstituted yeast in vitro translation assays (Saini et al. 2009). An eIF5A function in elongation is consistent with identification of an eIF5A mutant as a suppressor of nonsense-mediated mRNA decay (Zuk and Jacobson 1998), which depends on translation elongation (Peltz et al. 1992; Beelman and Parker 1994).

A prokaryotic ortholog of eIF5A termed EF-P was recently shown to stimulate translation of polyproline (Doerfel et al. 2013; Hersch et al. 2013; Ude et al. 2013), and a similar function was identified for yeast eIF5A (Gutierrez et al. 2013). Partial inactivation of a temperature-sensitive eIF5A variant confers reduced expression of reporter genes or authentic yeast ORFs containing homopolyproline stretches (Gutierrez et al. 2013), leading to defects in fertility and polarized cell growth (Li et al. 2014). The requirement for eIF5A for synthesis of polyproline sequences was also observed in vitro and shown to be dependent on the hypusine modification in eIF5A. Directed hydroxyl radical probing experiments revealed eIF5A binds near the ribosomal E site with the hypusine residue in the vicinity of the acceptor stem of the P-site tRNA and the peptidyl-transferase center of the ribosome. Thus, eIF5A is proposed to insert its hypusine residue into the active site of the ribosome and promote peptide bond formation especially for poor substrates like polyproline (Gutierrez et al. 2013).

Ribosomal frameshifting

The yeast system has proven a valuable genetic tool for dissecting the elongation pathway and in particular the fidelity of protein synthesis in vivo. The stoichiometry of structural and catalytic proteins produced by Ty retrotransposons of yeast is determined by a “programmed” +1 translational frameshift. For both Ty1 and Ty3 retrotransposons, a +1 ribosomal frameshifting event establishes the levels of Gag vs. Gag–Pol fusion proteins. While the precise mechanism of ribosomal frameshifting differs for Ty1 and Ty3, in both cases frameshifting is triggered by a codon that is decoded by a rare tRNA (Belcourt and Farabaugh 1990; Farabaugh et al. 1993). The L-A double-stranded RNA virus of yeast likewise encodes both Gag and Pol proteins with the catalytic Pol synthesized as a Gag–Pol fusion by −1 ribosomal frameshifting (Dinman et al. 1991). Mutations predicted to impair recycling of eEF1A–GDP to eEF1A–GTP and thus slow A-site tRNA binding enhance +1 ribosomal frameshifting that occurs on the P site, whereas the fungal eEF2-specific antibiotic sordarin as well as mutations that impair eEF2 function specifically inhibit +1 frameshifting (Dinman and Kinzy 1997; Harger et al. 2001, 2002). In contrast, mutations thought to impair GTP hydrolysis by eEF1A increase the probability of −1 frameshifting (Dinman and Kinzy 1997; Harger et al. 2001). These results are consistent with the notion that +1 frameshifting occurs by miscoding in the P site, while −1 frameshifting occurs on ribosomes in which both the A and P sites are occupied by tRNAs (Harger et al. 2002).

Ribosomal frameshifting is also used to control the production of cellular proteins in yeast. Polyamine-regulated +1 ribosomal frameshifting on the OAZ1 mRNA governs the synthesis of antizyme, a regulator of polyamine synthesis in cells (Palanimurugan et al. 2004; Kurian et al. 2011), and +1 frameshifting is also used to synthesize Est3, a subunit of telomerase (Morris and Lundblad 1997; Taliaferro and Farabaugh 2007), and Abp140 (Trm140), a tRNA methyltransferase (Asakura et al. 1998; D’Silva et al. 2011; Noma et al. 2011). Whereas +1 frameshifting can be triggered when the A site contains a stop codon or a sense codon decoded by a rare tRNA, −1 frameshifting relies on a slippery sequence that allows repairing of the A- and P-site tRNAs with the mRNA and on a downstream secondary structure that impedes forward movement of the ribosome. Computational approaches to identify slippery sites and potential downstream pseudoknot structures predict that as many as 10% of yeast genes, including several genes controlling telomere maintenance (Table 3, Advani et al. 2013), contain a frameshift signal (Jacobs et al. 2007; Belew et al. 2008) and that frameshifting on these sites impacts mRNA levels by activating the nonsense-mediated decay pathway (Belew et al. 2011).

Actin bundling and nontranslation functions of eEF1A

eEF1A has activities outside its canonical role in translation elongation, including functions in several steps of viral life cycles, apoptosis, actin bundling, and others in metazoans (Mateyak and Kinzy 2010). The interaction of eEF1A with actin was first identified in the slime mold Dictyostelium discoideum (Yang et al. 1990). Using a combination of genetic and/or biochemical approaches, this ability of eEF1A to bind actin and promote actin bundling was demonstrated to be conserved in both budding (Munshi et al. 2001) and fission yeast (Suda et al. 1999). Subsequent genetic studies in S. cerevisiae enabled an analysis of the functional interaction between these two highly abundant proteins in vivo. Utilizing both overexpression and mutational analysis, the actin-bundling activity of eEF1A was found to reside in multiple regions of the protein, in particular the N terminus and the C-terminal domain. Overexpression of eEF1A results in altered cell size, disorganization of the actin cytoskeleton, and accumulation of cells in the G1 phase of the cell cycle (Munshi et al. 2001), while a genetic screen designed to capitalize on these phenotypes identified a series of mutations in eEF1A that when coupled with an altered C terminus result in similar cellular effects (Gross and Kinzy 2005, 2007). The eEF1A mutant proteins, in particular the F309L and S405P mutants, reduced actin bundling in vitro, supporting the original biochemical data from the Condeelis laboratory (Yang et al. 1990), and caused defects in the cell cytoskeleton and morphology, revealing a critical biological role for this interaction. Most surprisingly, these eEF1A mutants showed translation defects at the level of initiation and increased phosphorylation of eIF2α (Perez and Kinzy 2014). Deletion of the eIF2 kinase GCN2 eliminated the initiation defect, but revealed an elongation defect. Taken together, these studies reveal linkages between the cellular cytoskeleton and translation and are consistent with early work highlighting this important interaction (Howe and Hershey 1984) .

Termination and recycling

Termination and ribosome recycling are linked processes critical to release the completed polypeptide and to provide a pool of 40S and 60S subunits for additional rounds of translation. Termination begins with recognition of one of the 3 stop codons in the A site by the release factor eRF1 (Sup45), which binds to the ribosome together with the GTP-bound form of factor eRF3 (Sup35). The eRF1, composed of three domains, functionally mimics a tRNA with the N-terminal domain recognizing the stop codon (Bertram et al. 2000), the central domain with its methylated GGQ motif promotes hydrolysis of the peptidyl-tRNA bond (Heurgue-Hamard et al. 2005), and the C-terminal domain interacts with eRF3. Recognition of all three stop codons by eRF1 is mediated by the YxCxxxF and TASNIKS motifs, as well as by other binding pockets/cavities in the N domain (Conard et al. 2012; Blanchet et al. 2015). The interaction between eRF1 and eRF3 is critical for stop codon recognition (Wada and Ito 2014), and GTP hydrolysis by eRF3 facilitates eRF1 discrimination of stop codons (Salas-Marco and Bedwell 2004) and accelerates peptide release (Eyler et al. 2013). Upon GTP hydrolysis eRF3 dissociates, leaving eRF1 in the A site. Binding of Rli1, an ABC-family ATPase, promotes eRF1-mediated hydrolysis of the aminoacyl bond linking the polypeptide to the peptidyl-tRNA (reviewed in Dever and Green 2012). Structural analysis of yeast eRF1 in complex with eRF3 or Rli1 on the ribosome revealed conformational changes in eRF1, while eRF3 and Rli1 were bound to the intersubunit space overlapping the binding sites for eEF1A and eEF2 (Preis et al. 2014). In complex with eRF3, eRF1 interacts with the stop codon; while in complex with Rli1, eRF1 no longer interacts with the stop codon but the GGQ motif is positioned toward the peptidyl-transferase center to promote hydrolysis. These structural changes reveal an uncoupling of stop codon recognition from peptide release by eRF1 (Preis et al. 2014). As would be expected, both eRF1 and eRF3 are essential for yeast viability; however, the N terminus of eRF3 can be deleted. This N-terminal prion domain of eRF3 is the basis of the [PSI⁺] aggregation of eRF3 (reviewed in Liebman and Chernoff 2012), resulting in impaired translation termination (Baudin-Baillieu et al. 2014).

Following release of the completed polypeptide, an 80S ribosome is bound to the mRNA with a deacylated tRNA in the P site base paired to the penultimate codon. ATP hydrolysis by Rli1 promotes release of the 60S subunit (Shoemaker and Green 2011). Depletion of Rli1 leads to aberrant reinitiation near the stop codon, leading to translation of 3′ UTR sequences (Young et al. 2015). In mammals, the protein Ligatin, or the complex of MCT-1 and DENR, promote release of mRNA and deacylated tRNA from the 40S subunit in the final step of recycling (Skabkin et al. 2010). The yeast orthologs of these 40S recycling factors [termed Tma64 (Ligatin), Tma20 (MCT-1), and Tma22 (DENR)] were previously identified as ribosome-associated proteins, and expression of human MCT-1 complemented translation defects in a strain lacking TMA20 (Fleischer et al. 2006). These results suggest that the mechanism of ribosome recycling is well conserved between yeast and mammals.

Global approaches to studying translation controls

Global translational activity in cell populations is often monitored by polysome profile analysis. “Freezing” ribosomes on mRNAs with CHX immediately prior to cell harvest provides a snapshot of translation. CHX binds to the ribosome E site, preventing tRNA release and trapping ribosome–mRNA complexes (Schneider-Poetsch et al. 2010). In recent years, formaldehyde cross-linking has also been used to preserve factor–ribosome interactions (Valasek et al. 2007), while for high-throughput sequencing applications, adding CHX at very high concentrations or only during cell lysis is now favored to avoid possible artifacts (Gerashchenko and Gladyshev 2014).

For polysome profile analyses, lysates are sedimented through sucrose gradients and then fractionated to generate an absorbance trace of rRNA that reveals a snapshot of the ribosome distribution (Pospisek and Valasek 2013). In actively growing cells, most ribosomes (85 ± 5%) are engaged in translation. mRNAs have a mean density of 0.64 ± 0.31 ribosomes per 100 nt coding region, with the density varying from <0.1 to >1.6 between mRNAs. Typically, longer ORFs have lower ribosome density (Arava et al. 2003). Stressed cells, or those with mutations that inhibit global translation initiation, exhibit an increased proportion of 80S ribosomes [monosomes and 80S couples (associated 40S and 60S subunits not bound to an mRNA)] relative to polysomes. Northern blotting or RT-PCR analysis of polysome gradient fractions have been used to study specific candidate genes, while microarray and high-throughput sequencing approaches such as ribosome footprint profiling (ribo-seq) have been used to assess changes in translation across multiple mRNAs simultaneously (Ingolia et al. 2009).

Stresses found to deplete bulk polysomes include the sudden withdrawal of glucose (Ashe et al. 2000; Arribere et al. 2011; Vaidyanathan et al. 2014) or amino acids (Smirnova et al. 2005); nutrient limitation that induces sporulation (Brar et al. 2012); temperature shift (Preiss et al. 2003); and the addition of cellular stress agents: hyperosmotic salt (Melamed et al. 2008), hydrogen peroxide (Shenton et al. 2006), fusel alcohols (Smirnova et al. 2005), or drugs: rapamycin (Preiss et al. 2003), calcofluor-white (Halbeisen and Gerber 2009), and chlorpromazine (Deloche et al. 2004). These studies have shown that there is widespread reprogramming of translation following stress with diminished ribosome association of some mRNAs hypersensitive to stress, while the translation of other mRNAs is relatively resistant to the stress. Moreover, the mechanisms of translational reprogramming are stress specific. For example, the global response to amino acid starvation is dependent on Gcn2 and eIF2α phosphorylation, while the response to hydrogen peroxide involves inhibiting both initiation via Gcn2 and translation elongation (Shenton et al. 2006), causing pausing at aspartic acid and serine codons according to ribo-seq experiments (Pelechano et al. 2015); however, antioxidant-response mRNAs become well translated (Shenton et al. 2006; Kershaw et al. 2015). The response to glucose depletion is rapid and does not require eIF2α phosphorylation, with almost all ribosomes being lost from mRNAs within 1–2 min (Ashe et al. 2000), perhaps via alterations in eIF4A function (Castelli et al. 2011).

Unexpectedly, some mRNAs, translationally activated 15 min after glucose withdrawal, share the same promoter sequence that binds Hsf1 (Zid and O’Shea 2014), suggesting that transcription and mRNA nuclear history might contribute to active translation during stress (Zid and O’Shea 2014). The fate of translationally repressed mRNAs following glucose depletion has also been analyzed. mRNAs enter cytoplasmic foci termed P bodies or stress granules or are degraded (Hoyle et al. 2007; Buchan et al. 2008; Arribere et al. 2011). P bodies and stress granules form with different kinetics and contain different protein components (Hoyle et al. 2007; Buchan et al. 2008) and repressed mRNAs enter P bodies at different times after stress (Simpson et al. 2014). After several hours of starvation, growth resumes with a distinct pattern of highly translated mRNAs, allowing enhanced synthesis of mitochondrial-targeted proteins (Vaidyanathan et al. 2014). Altered protein kinase A signaling contributes to the translational reprograming (Ashe et al. 2000; Tudisca et al. 2012; Vaidyanathan et al. 2014). Hence diverse cellular stress conditions can rapidly perturb global and mRNA-specific protein synthesis via multiple mechanisms to achieve stress-specific translational reprogramming. In addition to revealing translation changes under stress conditions, ribosomal profiling techniques have provided mechanistic insights into elongation through the use of distinct inhibitors (Lareau et al. 2014) as well as by identifying specific pauses during elongation (Pelechano et al. 2015).

Regulating eIF4E–eIF4G interactions by 4E-BPs

In higher eukaryotes, 4E-BPs play a prominent role in controlling eIF4E function and cellular translation (Richter and Sonenberg 2005). The 4E-BPs are a diverse set of proteins that share a common, albeit rather degenerate, motif YxxxxLΦ (where x is any residue and Φ is hydrophobic) that enables them to compete with eIF4G for binding to the surface of the cap-binding protein eIF4E. The 4E-BPs regulate a variety of cellular and developmental processes in higher eukaryotes (Kong and Lasko 2012). Yeast has two characterized 4E-BPs that contain the consensus eIF4E-binding motif, Caf20 (also called p20) and Eap1, which are 18 and 70 kDa, respectively (Altmann et al. 1997; Cosentino et al. 2000). Caf20 and Eap1 share no sequence similarity outside the eIF4E-binding motif, mutation of which abrogates eIF4E interactions (Altmann et al. 1997; Cosentino et al. 2000; Ibrahimo et al. 2006). Because mutations in eIF4E have differential impacts on the binding of eIF4G and Caf20, the eIF4E-interaction interfaces with eIF4G and Caf20 are likely overlapping, but distinct (Ptushkina et al. 1998). This potentially enables Caf20 to displace eIF4G from eIF4E. Recent structural studies show higher eukaryote 4E-BPs share similar eIF4E-binding properties (Peter et al. 2015). These data suggest the yeast proteins are parallels of the higher eukaryote 4E-BPs.

Both 4E-BPs are nonessential and deletion mutants display normal polysome profiles in optimum growth conditions (Cridge et al. 2010). The 4E-BPs play nonredundant roles in the adaptive growth response to nitrogen limitation as deletion mutants prevent pseudohyphal development and invasive growth of the Σ1278b strain (Ibrahimo et al. 2006), while S288c deletion strains are sensitive to growth on alternative nitrogen sources (Cridge et al. 2010). Translational repression upon cell treatment with the antipsychotic drug chlorpromazine or the oxidants cadmium and diamide are partially dependent upon Eap1 (Deloche et al. 2004; Mascarenhas et al. 2008), while caf20∆ displays synthetic growth phenotypes when combined with tif3∆ (eIF4B deletion) (de la Cruz et al. 1997) or the tif4631-459 allele that disrupts eIF4G1 binding to eIF4E (Hershey et al. 1999). These phenotypes are consistent with both 4E-BPs being translational repressors.

Targeted and global studies have been used to identify mRNA targets of Caf20 and Eap1 regulation. For example, the expression of STE12, GPA2, and CLN1 was shown to be translationally upregulated in caf20∆ cells during filamentous growth, and STE12 regulation was dependent on Caf20 (Park et al. 2006b). Eap1 (together with Scp160, Asc1, and Smy2) is involved in translational repression of the POM34 mRNA (Sezen et al. 2009). These data support the idea that the 4E-BPs can interact with and repress the translation of particular mRNA targets, possibly via interactions with other sequence-specific mRNA binding proteins, similar to examples from higher eukaryotes (Kong and Lasko 2012).

Microarray analysis of polysome-associated mRNAs identified >1000 genes whose polysome association was affected by deletion of either 4E-BP (Cridge et al. 2010). A computational analysis suggested that Caf20 binds mRNAs with structured 5′ UTRs (Cawley and Warwicker 2012). High-throughput sequencing of mRNAs associated with TAP-tagged Caf20 or Eap1 revealed that the two 4E-BPs bind mainly the same set (>1000) of longer than average mRNAs, suggesting that translation of these mRNAs is dampened or that their translation is poised for repression by 4E-BP binding (Costello et al. 2015). Most 4E-BP-bound mRNAs were also enriched in eIF4E interaction, as expected if the 4E-BPs repress translation via their interaction with eIF4E. However, some 4E-BP-bound mRNAs were not enriched for eIF4E, suggesting that the 4E-BPs can act independently of eIF4E (Costello et al. 2015). Over 100 mRNAs were found to interact with Caf20 independently of its ability to bind eIF4E (Castelli et al. 2015). The 3′ UTR of one mRNA tested (ERS1) directed Caf20-mediated repression of translation, and this regulation could be transplanted to a heterologous reporter (Castelli et al. 2015). In summary, both 4E-BPs can compete with eIF4G for binding to eIF4E on many mRNAs, resulting in impaired translational efficiency. However, the 4E-BPs may also act as repressors independently of eIF4E on some mRNAs.

Translationally regulated mRNAs

There are many examples of translationally controlled mRNAs in yeast. General features of translational control at the initiation phase include regulated recognition of uORFs by scanning ribosomes (GCN4), inhibition of scanning by mRNA secondary structure (HAC1), leaky scanning to initiate internally (MOD5 and GRS1), or translational repression during mRNA localization through recognition of mRNA structural elements by RNA-binding proteins (ASH1). In addition, the elongation and termination phases of translation can be regulated. For example, ribosome stalling (CPA1) as well as frameshifting (retrotransposons, L-A virus, and OAZ1) and stop codon read-through can lead to the production of alternate proteins. Programmed frameshifting can further alter the mRNA levels by triggering decay such as for EST1, EST2, STN1, and CDC13 (Advani et al. 2013).

In addition to the cap-dependent scanning model of translation initiation, an alternative cap-independent mode of translation has been proposed for some mRNAs in higher organisms (Jackson 2013). While this alternate mode of translation initiation is best characterized for viral mRNAs, some cellular mRNAs may also employ this alternate means of translation initiation (Gilbert 2010; Jackson 2013). In general, cap-independent initiation is thought to rely on special secondary structure elements in the mRNAs to recruit translation factors and/or the 40S subunit (or PIC). In experiments employing in vitro translation assays and mRNA electroporation experiments, the translation of several yeast mRNAs encoding proteins required for invasive growth was maintained when the canonical m⁷GTP cap was replaced by an ApppG cap structure (Gilbert et al. 2007). This cap-independent translation was dependent on eIF4G and was attributed to a poly(A) element in the 5′ UTR of the mRNAs (Gilbert et al. 2007). Similarly, the URE2 mRNA was reported to direct the synthesis of both full-length Ure2 and an N-terminally truncated Ure2 that initiates from an internal AUG codon (Komar et al. 2003). Interestingly, synthesis of the truncated Ure2 was maintained upon genetic inactivation of the mRNA cap-binding protein eIF4E and when sequences with high secondary structure were inserted in the 5′ UTR to block scanning ribosomes (Komar et al. 2003). This putative internal initiation of translation on the URE2 mRNA was enhanced in cells lacking the protein eIF2A (YGR054w) (Komar et al. 2005) and was found to depend on a 104-nt A-rich stem-loop element that includes the internal AUG start codon (Reineke et al. 2008; Reineke and Merrick 2009). Currently it is unclear under what conditions cap-independent translation will be important in yeast and whether the mRNA elements that support cap-independent translation might function as general translational “enhancers” to help the mRNAs compete for the translational apparatus by the conventional cap-dependent pathway (Gilbert 2010).

Selected examples of translationally controlled mRNAs and associated references are given in Table 3. Detailed discussions of GCN4 and CPA1 control mechanisms are presented below.

GCN4

Probably the best-characterized example of regulation of protein synthesis in yeast is GCN4. Environmental signals are transduced to modulate start codon selection on the GCN4 mRNA and limit production of the Gcn4 protein to specific cellular stress conditions. Gcn4 is a basic leucine zipper (bZIP) transcriptional activator of amino acid and related biosynthetic genes (Natarajan et al. 2001). GCN4 mRNA translation is controlled by a reinitiation mechanism that requires an interplay of sequences in its 5′ leader including uORFs with translation initiation factors and ribosomes. Under nonstarvation conditions the flow of ribosomes to the GCN4 AUG start codon is limited by up to ∼100-fold and GCN4 translation is repressed unless cells are starved (Hinnebusch 2005).

Following starvation for one or more amino acids, a signaling pathway, termed general amino acid control, is deployed, which activates Gcn2 to phosphorylate eIF2α (Dever et al. 1992) causing inhibition of eIF2B (Pavitt et al. 1998) and a reduction in eIF2 TC levels (Dever et al. 1995). The lower level of TC leads to reduced ribosome engagement with most mRNAs; however, paradoxically, more ribosomes reach the GCN4 AUG codon and Gcn4 levels increase by up to 10-fold (Albrecht et al. 1998). Because GCN4 translation is acutely sensitive to the levels of active TC, it has proved an extremely useful tool to probe the role of translation factors in the scanning mechanism of protein synthesis as well as many of the details of translational control by eIF2 kinases (Hinnebusch 2005; Hinnebusch 2011).

Arginine-regulated ribosome stalling controls CPA1 translation

CPA1 encodes the small subunit of carbamoyl phosphate synthetase, an enzyme that catalyzes a step in the synthesis of citrulline, an intermediate in the arginine biosynthesis pathway. CPA1 mRNA translation is regulated by a uORF, and by Arg, via a ribosome-stalling mechanism that is conserved across fungi (Hood et al. 2007). The CPA1 uORF, YOR302W, encodes a 25 amino acid arginine attenuator peptide (AAP). The peptide sequence, especially residues 6–23, is critical for translational repression by Arg (Werner et al. 1987; Delbecq et al. 2000; Hood et al. 2007). Many single missense mutations in the uORF eliminate Arg-controlled ribosome stalling (Delbecq et al. 2000). Of these, the D13N mutant has been used widely to inform mechanistic understanding (Wang et al. 1999; Gaba et al. 2001). In contrast to GCN4, most nucleotide sequences around the CPA1 uORF are not important (Delbecq et al. 1994).

In the absence of Arg, both the uORF and CPA1 are translated. The uORF AUG context is poor, ensuring that some scanning PICs bypass the uORF AUG and instead initiate translation at CPA1 (Werner et al. 1987). This was confirmed by ribosome toe print analyses using a uORF-regulated luciferase reporter and yeast translation extracts (Gaba et al. 2001), as well as via ribo-seq experiments (Ingolia et al. 2009). There is no evidence supporting ribosome reinitiation after uORF translation. The presence of high Arg concentrations induces ribosome stalling at the uORF stop codon, with 80S complexes retaining a P-site tRNA-linked nascent peptide within the ribosome exit tunnel. The stalled 80S prevents any PICs that leaky scan through the uORF AUG codon from progressing to the CPA1 AUG codon (Wang et al. 1999; Gaba et al. 2001). This ensures Cpa1 levels drop when Arg is abundant.

CPA1 control is not mediated by monitoring tRNA Arg charging levels and the Saccharomyces AAP sequence does not contain Arg residues (Wang et al. 1999). Hence it is mechanistically distinct from both GCN4 uORF control and the ribosome stalling associated with Trp attenuation in bacteria. Studies performed using the orthologous Neurospora crassa Arg-2 locus have helped inform the mechanism of peptide and Arg-induced stalling. Control requires Arg itself, which alters interactions between the P-site tRNA/nascent AAP and both rRNA and ribosomal proteins within the PTC and the peptide exit tunnel of the 60S. Although ribosome stalling naturally occurs at the end of the AAP sequence, it was shown that Arg-dependent stalling occurs during translation elongation as removing the stop codon to extend the peptide or transferring the AAP sequence to the middle of a reporter gene generated novel Arg-mediated ribosome-stalling contexts (Wang et al. 1998; Fang et al. 2004). Structural analysis of the 80S-bound stalled nascent peptide by cryo-EM revealed that residues 10–24 of the N. crassa AAP (equivalent to residues 11–25 of yeast AAP) forms an α-helix within the exit tunnel and that AAP makes a series of contacts with tunnel-exposed conserved 28S rRNA bases in the upper tunnel and with residues of Rpl4 and Rpl17 at the exit tunnel constriction point (Bhushan et al. 2010). Complementary analyses with photo-cross-linked amino acids suggest that Arg alters the AAP conformation within the exit tunnel, affecting its interactions with both Rpl4 and Rpl17 (Wu et al. 2012). The cryo-EM data also suggest that Arg stabilizes a distinct conformation of 28S rRNA residue A2062 such that the AAP-linked P-site tRNA conformation within the PTC could prevent eRF1 action (Bhushan et al. 2010). An in vitro translation puromycin release assay confirmed that Arg-mediated stalling inhibits peptidyl transfer activity, thereby preventing normal translation termination and ribosome recycling (Wei et al. 2012). How Arg interferes with PTC function is not yet clear.

As stalling occurs at the uORF stop codon, Cpa1 levels are further controlled by the NMD quality control pathway that recognizes mRNAs with aberrant premature stop codons (Kervestin and Jacobson 2012). NMD requires three proteins: Upf1, Nmd2, and Upf3. When 80S complexes translating the uORF stall in the presence of Arg, the CPA1 mRNA is destabilized, dependent on Upf1 (Messenguy et al. 2002) and Nmd2 (Gaba et al. 2005). Consistent with these findings, mutating each NMD gene enhances Cpa1 activity in the presence of Arg (Messenguy et al. 2002). Translation of the AAP and ribosome stalling over the uORF stop codon are necessary for NMD as mutations altering the AUG start codon or the D13N mutation in the AAP eliminate NMD (Gaba et al. 2005).