Translational Regulation of Yeast GCN4

Disclosed is a pneumatic launch apparatus which employs separable telescoping tubes. The open ends of the tubes are provided with valves which automatically close upon separation of the tubes to trap operating gas in the tubes at the moment of separation.

When subjected to starvation, stress, or viral infections, mammalian cells down-regulate general protein synthesis by phosphorylating the ␣ subunit of eukaryotic translation initiation factor 2 (eIF2) 1 (1). eIF2 functions in translation initiation by delivering charged initiator tRNA Met (Met-tRNA i Met ) in a ternary complex with GTP to the 40 S ribosomal subunit, forming a 43 S preinitiation complex. In the translation of most mRNAs, the 43 S complex binds near the capped 5Ј end, migrates downstream, and upon reaching the first AUG codon, joins with the 60 S subunit to form an 80 S initiation complex (the scanning mechanism) (2). Following AUG recognition, the GTP bound to eIF2 is hydrolyzed and eIF2 is released as an inactive eIF2⅐GDP binary complex. Exchange of the GDP bound to eIF2 with GTP is catalyzed by eIF2B. Phosphorylation of the ␣ subunit of eIF2 (eIF2␣) on Ser-51 prevents the recycling of eIF2 by eIF2B; in addition, the phosphorylated complex eIF2(␣P)⅐GDP has a higher affinity than non-phosphorylated eIF2⅐GDP for eIF2B, such that GDP-GTP exchange on non-phosphorylated eIF2 is also impaired and ternary complex formation is blocked (2). In Saccharomyces cerevisiae, eIF2␣ is phosphorylated when cells are deprived of an amino acid or purine, and interestingly, this leads to increased translation of a specific mRNA encoding GCN4, a transcriptional activator of at least 40 genes encoding amino acid biosynthetic enzymes (3).
The unique induction of GCN4 translation in response to eIF2 phosphorylation is mediated by four short open reading frames (uORFs) in the leader of GCN4 mRNA located 150 -360 nucleotides upstream of the authentic initiation codon. Eliminating the start codons of all four uORFs results in high level GCN4 expression under both starvation and non-starvation conditions without altering the mRNA (4). Thus, the uORFs inhibit GCN4 translation in non-starved cells by restricting the progression of scanning ribosomes through the leader to the GCN4 start codon. The first and fourth uORFs (from the 5Ј end), which are sufficient for nearly wild-type regulation, have different effects on GCN4 translation. When present alone, uORF4 reduces GCN4 translation to only 1% of the level seen in the absence of all four uORFs, under both starvation and non-starvation conditions (4). In this situation, it appears that all ribosomes translate uORF4 and then dissociate from the mRNA. In contrast, uORF1 alone reduces GCN4 translation by only 50%, presumably because half of the ribosomes that translate uORF1 resume scanning and reinitiate at GCN4. When uORF1 is present upstream of uORF4 and amino acids are abundant, all the ribosomes that resume scanning following uORF1 translation will reinitiate at uORF4 and dissociate from the mRNA, preventing GCN4 translation. Under starvation conditions, however, half of the ribosomes that resume scanning after uORF1 translation will bypass the uORF4 start site (and also those at uORFs 2-3 when present) and reinitiate at GCN4 instead ( Fig.   1) (5). Thus, prior translation of uORF1 allows ribosomes to overcome the strong translational barrier at uORF4 by a reinitiation mechanism.
We proposed that under conditions where GCN4 is repressed, the ribosomes that resume scanning following uORF1 translation are forced to reinitiate at uORF4 because they rebind the eIF2⅐GTP⅐Met-tRNA i Met ternary complex before reaching the uORF4 start codon. Under starvation conditions, phosphorylation of eIF2␣ reduces the concentration of ternary complexes, such that many ribosomes scan the distance between uORF1 and uORF4 without rebinding the ternary complex (6). Lacking initiator tRNA Met , they cannot recognize the AUG codons at uORFs 2, 3, and 4 (7) and continue scanning downstream. Most of these ribosomes will bind the ternary complex while scanning between uORF4 and GCN4, enabling them to reinitiate at the GCN4 start codon (6). Thus, reducing the level of ternary complexes allows ribosomes to bypass the inhibitory uORFs 2-4 and reinitiate at GCN4 instead (6) (Fig. 1).

Genetic Evidence for the Scanning-Reinitiation Mechanism
of Translating GCN4 mRNA Supporting the idea that any ribosomes which translate GCN4 must have scanned past uORF4 without initiating translation ( Fig.  1), it was shown that translation of a uORF4-lacZ fusion decreases under conditions where GCN4 translation is stimulated (5). Another strong indication that uORF4 must be skipped en route to GCN4 is that mutations in the uORF4 stop codon that make uORF4 overlap the GCN4 ORF have almost no effect on GCN4 expression, indicating that the location of the uORF4 stop codon is of little consequence (5). In contrast, elongating uORF1 abolishes GCN4 translation (8), supporting the idea that essentially all ribosomes reach the GCN4 start site by reinitiation following translation of uORF1. In addition, there is a critical requirement for nucleotides flanking the uORF1 stop codon for efficient reinitiation at GCN4. Replacing the last codon and 10 nucleotides 3Ј to the uORF1 stop codon with the corresponding nucleotides from uORF4 converts uORF1 into a strong translational barrier and destroys its ability to stimulate GCN4 translation when situated upstream from uORF4 (9). Mutational analysis revealed that diverse AU-rich sequences at the third codon and immediately 3Ј of uORF1 would promote high level reinitiation at GCN4 (10). This led to the idea that base pairing between the mRNA surrounding the uORF4 stop codon and the rRNA could lengthen the time spent by the ribosome in the termination region and increase the probability of ribosome release from the mRNA.
It was conceivable that following termination at uORF1, ribosomes would be shunted directly to the GCN4 start site rather than scanning the entire uORF1-GCN4 interval. This possibility is inconsistent with the fact that insertions of stem-loop structures in the vicinity of uORF4 abolish GCN4 translation. A shunting model also cannot explain the critical finding that GCN4 translation gradually decreased as the uORF1-uORF4 spacing was progressively increased (5). According to the model in Fig. 1, under derepressing conditions, 50% of the 40 S subunits scanning from uORF1 have not re-bound the ternary complex upon reaching uORF4 and continue scanning to GCN4. When the uORF1-uORF4 interval is expanded, however, most of the 40 S subunits have bound the ternary complex and are now competent to reinitiate when they reach uORF4. Consequently, they cannot bypass uORF4 and reinitiate downstream at GCN4. A final piece of evidence inconsistent with a shunting mechanism is that the authentic uORFs were replaced with heterologous small uORFs without destroying GCN4 translational control (11)(12)(13). In our model, the only critical requirements for the first uORF are to be recognized efficiently by ribosomes and then to allow ribosomes to resume scanning. Presumably, both requirements can be met by heterologous uORFs, albeit with lower efficiency than occurs with authentic uORF1. Interestingly, efficient reinitiation at uORF1 depends not only on the sequence context of its stop codon but on sequences upstream of the uORF (14).
A key feature of the model in Fig. 1 is that the reduction in ternary complex levels in starved cells is large enough to allow a fraction of reinitiating ribosomes on GCN4 mRNA to ignore uORF4 and continue scanning to GCN4 but not extensive enough to allow ribosomes to skip uORF4 if they have not translated uORF1. This distinction may exist because in conventional initiation events ribosomes bind the ternary complex before interacting with the 5Ј end of the mRNA (2); therefore, reducing ternary complex levels may decrease the frequency that ribosomes load at the 5Ј end, but once bound to mRNA, their ability to recognize AUG codons while scanning downstream should be independent of the concentration of ternary complexes. It is also possible that reinitiating 40 S subunits are less efficient than free 40 S subunits in binding the ternary complex because they lack an initiation factor like eIF3 or eIF1A that promotes this reaction (2). In any case, the reinitiation mechanism for GCN4 translation is an extremely sensitive indicator of the activity of eIF2 and associated factors that function in the formation of ternary complex or promote its binding to ribosomes.

The Protein Kinase GCN2 Stimulates GCN4 Translation by
Inhibiting Recycling of eIF2 by eIF2B GCN2 is a 180-kDa protein kinase that phosphorylates eIF2 and thereby induces GCN4 translation in cells starved for histidine (6) or several other amino acids (15). Substitution of Ser-51 in eIF2␣ with Ala (the SUI2-S51A allele) completely eliminates the increased phosphorylation in amino acid-starved cells and impairs derepression of GCN4 to the same extent as a deletion of GCN2. Immunopurified GCN2 specifically phosphorylated the ␣ subunit of eIF2 purified from rabbit or yeast but not yeast eIF2 containing the Ala-51 substitution (6). These results established that GCN2 stimulates GCN4 translation by phosphorylating eIF2␣ on Ser-51. Low level expression of the mammalian eIF2␣ kinases HRI and PKR in gcn2 mutants induces GCN4 translation in a manner completely dependent on Ser-51 in eIF2␣ (16). When expressed at high levels, PKR and HRI produce a much higher level of eIF2 phosphorylation than occurs when GCN2 is activated in amino acid-starved cells, and this severely inhibits cell growth (16,17). Mutationally activated forms of GCN2 (GCN2 c kinases) also cause hyperphosphorylation of eIF2 and a general reduction in translation initiation (6, 18 -20). These last findings confirm that GCN4 translation is in-duced at a lower level of eIF2 phosphorylation than is required for general inhibition of protein synthesis.
There is strong evidence that phosphorylation of eIF2␣ by GCN2 down-regulates the formation of eIF2⅐GTP⅐Met-tRNA i Met ternary complexes. Mutations in the genes encoding the ␣, ␤, and ␥ subunits of eIF2 (encoded by SUI2 (21), SUI3 (22), and GCD11 (23), respectively) mimic the effect of eIF2 phosphorylation in derepressing GCN4 translation (Gcd Ϫ phenotype) independently of GCN2 and amino acid starvation (24 -27). These mutations produce a slow growth phenotype (Slg Ϫ ) on nutrient-rich medium and thus appear to decrease eIF2 function in translation initiation. Deleting two of the four IMT genes encoding tRNA i Met elicits the same Gcd Ϫ and Slg Ϫ phenotypes (28). Overexpression of the eIF2 complex in wild-type cells interferes with derepression of GCN4 translation and suppresses the growth inhibitory effects of GCN2 c alleles. Overexpressing eIF2 and tRNA i Met together has a synergistic effect in suppressing the toxicity of eIF2 hyperphosphorylation (28). The ␤, ␥, ␦, and ⑀ subunits of yeast eIF2B (encoded by GCD7 (29), GCD1 (30), GCD2 (31), and GCD6 (29), respectively) were first identified by point mutations with the same Gcd Ϫ and Slg Ϫ phenotypes observed for mutations in eIF2 subunits (24,(32)(33)(34)(35)(36)(37)(38). The fact that non-lethal mutations in the essential subunits of eIF2B mimic eIF2 phosphorylation in derepressing GCN4 supports the notion that eIF2 phosphorylation in yeast leads to a reduction in eIF2B function. Even stronger evidence for this conclusion comes from the fact that overexpressing the eIF2B complex in yeast overcomes the growth inhibitory effects and the derepression of GCN4 associated with eIF2 phosphorylation (28).
Mutations were obtained in the GCD2, GCD7, and GCN3 subunits of eIF2B that reverse the derepression of GCN4 and general inhibition of translation in the presence of high level eIF2 phosphorylation. These mutations appear to make eIF2B insensitive to eIF2(␣P) without decreasing the ability to catalyze nucleotide exchange on non-phosphorylated eIF2 (16, 39 -41). Interestingly, this is the only effect observed when GCN3 is deleted (16,39), indicating that GCN3 is required primarily for inhibition of eIF2B by eIF2(␣P). The same conclusion was recently obtained for rat eIF2B by showing that a four-subunit complex lacking the ␣-subunit (homologous to GCN3), reconstituted in baculovirus-infected insect cells, was insensitive to inhibition by phosphorylated eIF2 in vitro (42). The GCD2 and GCD7 subunits are essential (29, 36) and thus make important contributions to catalysis as well as regulation. GCN3, GCD7, and GCD2 show strong sequence similarities to one another throughout most of GCN3 and GCD7 and the C-terminal half of GCD2 (29,31). Overexpression of GCD2, GCD7, and GCN3 reduces the inhibitory effect of eIF2(␣P) on general translation in vivo, and the excess amounts of these proteins form a stable subcomplex that can be co-immunoprecipitated from cell extracts. Formation of this subcomplex does not compensate for a loss of eIF2B function by mutation; thus the GCN3⅐GCD7⅐GCD2 subcomplex does not possess guanine nucleotide exchange activity but instead appears to prevent eIF2(␣P) from inhibiting native eIF2B (43). Together these results provide strong evidence that GCN3, GCD7, and the C-terminal half of GCD2 comprise a regulatory domain in eIF2B that mediates the inhibitory effects of eIF2(␣P).
The majority of the regulatory mutations isolated in GCD2, GCD7, and GCN3, which decrease or abolish inhibition of eIF2B by eIF2(␣P), fall into two clusters of 70 amino acids in regions of strong similarity among all three proteins, and several mutations alter equivalent positions in two or all three subunits. These results suggest that structurally related segments in GCD2, GCD7, and GCN3 carry out similar roles in the regulation of eIF2B by eIF2(␣P). One possible function would be to interact with residues in eIF2␣ surrounding Ser-51 to increase the affinity of eIF2B for phosphorylated eIF2; alternatively they could elicit a conformational change in the eIF2⅐eIF2B complex that prevents nucleotide exchange on eIF2(␣P) (41). In certain GCD2 regulatory mutants, nearly all of the eIF2 was phosphorylated with no effects on cell growth. These mutations probably allow eIF2B to accept phosphorylated eIF2 as a substrate rather than simply reducing the affinity for eIF2(␣P) (41). Given the role of GCD2, GCD7, and GCN3 in the regulation of eIF2B, we proposed that GCD1 and GCD6 interact with one another and form the catalytic center for nucleotide exchange (44). This proposal is supported by recent findings that the rat homologue of GCD6 expressed in baculovirus-infected insect cells possessed low level exchange activity that was greatly stimulated by coexpressing the other four subunits in the same cells (42). The ␥ subunit of eIF2 (encoded by GCD11) is homologous to EF-Tu and contains sequences conserved in all known GTP-binding proteins (23). Moreover, several GCD11 mutations that reduce eIF2 function in vivo map within the putative GTP-binding domain (45). Thus, it seems likely that GCD11 contains the GTP-binding site on eIF2 and consequently would be expected to interact directly with the catalytic center in eIF2B.
Mutations in the GCD10 gene cause derepression of GCN4 translation in the absence of eIF2 phosphorylation by GCN2 and confer temperature-sensitive growth on rich medium (24,25). These are the same phenotypes associated with gcd mutations affecting subunits of eIF2 or eIF2B. A reduction in polysome content observed in gcd10 mutants at the non-permissive temperature is indicative of a general defect in translation initiation, and a small fraction of GCD10 protein was found associated with polysomes and ribosomal subunits. GCD10 co-purified with eIF3 biochemical activity and was shown to be identical to the RNA-binding 62-kDa subunit of eIF3 (46) described previously (47). These findings suggested that some aspect of translational reinitiation on GCN4 mRNA involves the GCD10-encoded subunit of eIF3. Based on the known biochemical activities of mammalian eIF3 (2), one possibility is that gcd10 mutations decrease the ability of eIF3 to stimulate formation of eIF2⅐GTP⅐Met-tRNA i Met ternary complexes or promote their binding to ribosomal subunits.

Regulation of GCN2 Kinase Activity by Uncharged tRNA
When GCN2 was replaced by the mammalian eIF2␣ kinases PKR or HRI, GCN4 translation was stimulated independently of amino acid levels (16), indicating that increased phosphorylation of eIF2␣ in starved cells reflects increased GCN2 activity, not inhibition of an eIF2␣ phosphatase. There is little or no increase in GCN2 protein levels in response to amino acid starvation (19), showing that GCN2 function, not its expression, is stimulated in starved cells. Uncharged tRNA appears to be the activating ligand for GCN2 because mutations in aminoacyl-tRNA synthetases lead to increased eIF2␣ phosphorylation by GCN2 (15) with attendant derepression of GCN4 (48) and genes under its control (15,49,50) without limitation for the cognate amino acids. GCN2 contains 530 residues C-terminal to the kinase domain similar to the sequence of histidyl-tRNA synthetase (HisRS) (51), including the conserved "motif 2" sequence that interacts with the acceptor stem of tRNA in class II synthetases (52). Accordingly, it was proposed that binding of uncharged tRNA to the HisRS-like domain stimulates the ability of the kinase domain to phosphorylate eIF2␣ (51). Supporting this model, it was shown that motif 2 in GCN2 is required for kinase function in vivo and in vitro and for tRNA binding by the HisRSlike domain in vitro (15,61). In addition, numerous activating GCN2 c mutations alter residues in the HisRS-like region (18 -20). GCN2 can interact with translating ribosomes and free ribosomal subunits, and this association is dependent on its C-terminal 120 amino acids (53). Interestingly, GCN2 c alleles that activate GCN2 most effectively alter residues in this C-terminal domain (18 -20), leading to the suggestion that activation of GCN2 occurs on the ribosome by uncharged tRNA bound to the ribosomal acceptor (A) site (18). This would be akin to the activation of the RelA protein of Escherichia coli by uncharged tRNA in the stringent response (54) (Fig. 2).
It has been proposed that GCN2 binds specifically to GCN4 mRNA, and this interaction is required for efficient derepression of GCN4 translation. Presumably, binding to GCN4 mRNA would cause GCN2 to phosphorylate eIF2 in a localized manner and stimulate GCN4 translation without affecting other mRNAs (55).

FIG. 2.
A model describing an EF3-related function of GCN1 and GCN20 on translating ribosomes in activation of protein kinase GCN2 by uncharged tRNA. A, right arrow, EF3 stimulates binding of cognate aminoacyl-tRNA⅐GTP⅐EF1␣ complexes to the ribosomal A site at the expense of non-cognate tRNA binding, consistent with its proposed role in translational fidelity. Left arrow, EF3 stimulates release of deacylated tRNA from the ribosomal E site. B, in E. coli, limitation for an amino acid (aa) results in accumulation of uncharged tRNA in the cytosol. RelA interacts with ribosomes in the vicinity of the A site, and binding of uncharged tRNA to the A site stimulates the production of ppGpp by RelA. Increased levels of ppGpp trigger the stringent response. C, GCN2 is a ribosome-associated eIF2␣ kinase with a HisRS domain that binds tRNA in vitro and is required for kinase activation by uncharged tRNA in amino acid-starved cells. By analogy with RelA in bacteria and considering the importance of the EF3like region in GCN1, we propose that the GCN1-GCN20 complex binds near the A site and functions in an EF3-like manner to stimulate the binding of uncharged tRNA to the A site or its delivery to the HisRS-like domain in GCN2. PK, protein kinase; ABCs, ATP-binding cassettes. E, P, and A, E, P, and A sites of the ribosome.
While this mechanism may increase the sensitivity of GCN4 mRNA to eIF2 phosphorylation by GCN2, it cannot account for the fact that mutations in eIF2 or eIF2B subunits lead to high level GCN4 translation with little effect on other mRNAs or on GCN4 derivatives lacking uORFs (24,25). It still seems necessary to postulate that reinitiation on GCN4 mRNA is much more sensitive than are conventional initiation events to reductions in ternary complex levels.
The GCN1-and GCN20-encoded proteins are also required for activation of GCN2 in starved cells. Mutations that inactivate these proteins either reduce (GCN20) or abolish (GCN1) phosphorylation of eIF2␣ by GCN2 but have no such effect in strains expressing PKR in place of GCN2 (56,57). The latter indicates that GCN1 and GCN20 are required to increase GCN2 kinase function rather than inhibit an eIF2␣ phosphatase. Neither gcn1 nor gcn20 mutations reduce GCN2 expression or decrease GCN2 kinase activity in immune complex assays (56,57), suggesting that GCN1 and GCN20 are required to mediate activation of GCN2 in vivo by uncharged tRNA. GCN1 and GCN20 are physically associated with one another in cell extracts (57), and substantial amounts of both proteins cosediment with polysomes and 80 S ribosomes (58). GCN1 and GCN20 share sequence similarity with translation elongation factor EF3 (56,57), an essential protein that stimulates binding of the EF1␣⅐GTP⅐aminoacyl-tRNA ternary complex to the A site and release of deacylated tRNA from the ribosomal E (exit) site (59). GCN20 contains two domains highly related to the ATP binding cassettes (ABCs) in EF3 but shows little sequence similarity to EF3 outside of these regions (57), whereas GCN1 is most similar to the region in EF3 N-terminal to the ABCs (56). The N-terminal 15-25% of GCN20, which is critically required for its regulatory function, interacts with the internal segment of GCN1 that is similar to EF3, and the EF3-like domain was found to be highly conserved in a human homologue of GCN1. Based on these findings, it was suggested that the GCN1⅐GCN20 complex functions on the ribosome in an EF3-like fashion to mediate activation of GCN2 by uncharged tRNA (Fig. 2) (58). A Drosophila homologue of GCN2 was recently discovered whose expression is developmentally regulated (60). This and the fact that proteins highly related to GCN1 (58) and GCN20 (57) appear to be present in human cells suggest that the GCN2⅐GCN1⅐GCN20 apparatus for modulating eIF2 function in translation initiation may be conserved in all eukaryotes and play important roles in regulating cell growth and differentiation.