The Unit on Protein Biosynthesis is studying how the function of eukaryotic translation initiation factor 2 (eIF2) is regulated by phosphorylation, with particular emphasis on determinants of substrate recognition by the eIF2a protein kinases and the structure and regulation of these enzymes. By delivering tRNAiMet to the ribosome, eIF2 plays a pivotal role in the initiation of protein synthesis; the factor also functions in recognition of the AUG start codon. Down-regulation of eIF2 by phosphorylation of its a-subunit is a highly conserved mechanism for inhibiting general protein synthesis under starvation and stress conditions in mammalian cells, and it mediates gene-specific translational induction of GCN4 expression in yeast. The human eIF2a kinase PKR has growth-suppressive properties, and its induction is an important aspect of the interferon-mediated antiviral response. As a consequence, many viruses encode antagonists of PKR. We have reconstituted in yeast cells the antagonism of PKR by pseudosubstrate inhibitors encoded by vaccinia and swine pox virus. By identifying mutations in these inhibitory proteins, and others in PKR, which increase or decrease inhibition, we are identifying molecular determinants of substrate recognition in both eIF2a and PKR.

We are also analyzing the function in translation initiation of the recently discovered factor eIF5B. An ortholog of the bacterial translation initiation factor IF2, eIF5B is universally conserved as indicated by the identification of homologs in yeasts, humans, and archea. We previously showed that eIF5B (encoded by the FUN12 gene in yeast) functions in ribosomal subunit joining and constitutes a novel, second GTP requirement in the translation initiation pathway. Recently, we obtained the x-ray structure of eIF5B, and we are studying the structure/function properties of this factor and its interaction with the ribosome by using a combination of in vivo and in vitro assays.

Molecular, Biochemical, and Structural Analysis of eIF5B/IF2, a Universally Conserved Translation Initiation Factor
Arefin, Cao, Shin, Dever in collaboration with Burley,a Pestovab
Our previous studies on the yeast FUN12 protein and homologs that we identified in both the human and archea established that these proteins are translation initiation factors, which we have named eIF5B. Interestingly, eIF5B is an ortholog of the bacterial translation initiation factor IF2. Subsequently, working in collaboration with Tatyana Pestova and Christopher Hellen at SUNY-Brooklyn, we found that the human eIF5B was required for the subunit joining step of protein synthesis. Like bacterial IF2, eIF5B binds GTP and ribosomes and displays ribosome-dependent GTPase activity (Pestova et al., 2000a; 2000b).

To gain further insights into the structure/function properties of this universally conserved translation initiation factor, we collaborated with Stephen Burley to determine the structure of eIF5B. The x-ray structure of eIF5B from the archeon Methanobacterium thermoautotrophicum was determined in three states: inactive (eIF5BoGDP), active [eIF5BoGDPNP (mimicking GTP)], and nucleotide-free. The extended protein is composed of four domains and resembles a chalice. The GTP-binding domain (domain I), a b barrel domain II, and a novel a/b/a-sandwich domain III constitute the cup of the chalice. A long a helix is connected to the cup of the chalice and forms the stem. The stem, in turn, connects to a second b barrel domain (IV), which forms the base of the chalice. Comparison of the structures of the active and inactive forms of eIF5B reveal that modest structural changes in the GTP-binding domain are amplified by an articulated lever mechanism resulting in significant movement of domain IV. The GTP-binding domain and domain II of eIF5B resemble the first two domains of the translation elongation factors EF-Tu and EF-G. Based on this structural conservation and on similar surface charge properties for the GTP-binding domains of eIF5B and EF-Tu, we propose that eIF5B binds near the GTPase-activating center of the ribosome, as has been shown for EF-Tu and EF-G (Roll-Mecak et al., 2000).

To assess the importance of GTP binding and hydrolysis by eIF5B, mutations were introduced at several residues conserved among all GTP-binding proteins. Mutations that eliminated the ribosome-dependent GTPase activity of eIF5B likewise prevented complementation of the slow-growth phenotype in yeast cells lacking eIF5B. In addition, recombinant forms of the eIF5B GTP-binding domain mutants failed to restore translation of a reporter mRNA in extracts prepared from a strain lacking eIF5B. Correlation of these in vivo and in vitro results indicates that GTP hydrolysis by eIF5B is critical for the translational stimulatory activity of the factor. In wild-type yeast cells, alternative reinitiation at upstream open reading frames in the GCN4 mRNA leader regulates ribosome access to the GCN4 start site. Using a set of GCN4 alleles with altered translational leader sequences, we have obtained evidence that lack of eIF5B causes an increase in leaky scanning wherein ribosomes scan past the first AUG codon on the GCN4 mRNA without initiating translation. The increase in leaky scanning provides in vivo evidence in support of the requirement of eIF5B for ribosomal subunit joining.

The identification of eIF5B was particularly intriguing because it was widely accepted that only a single GTP molecule was used during translation initiation. However, the translation initiation factor eIF2 is also known to be a GTPase; therefore, the discovery of eIF5B predicts that a second GTP molecule will be consumed during translation initiation. To test this prediction, a mutant form of human eIF5B was generated, changing the nucleotide specificity of the factor from GTP to XTP. Ribosome-dependent nucleotide hydrolysis assays confirmed that the mutation eliminated GTPase activity but endowed the factor with XTPase activity. Consistent with the altered nucleotide specificity, the mutant protein catalyzed subunit joining in the presence of XTP, but not GTP. Use of the mutant protein in translation initiation assays revealed an essential requirement for XTP, in addition to the expected GTP requirement of eIF2 for methionyl-tRNA binding. The results demonstrate that two GTP molecules are required in eukaryotic translation initiation. We propose that the GTP used by eIF2 plays a role in proper selection of the AUG start codon, whereas the GTP consumed by eIF5B promotes ribosomal subunit joining.

Based on the structure of the factor, two additional mutational analyses of eIF5B have been initiated. The Switch I is a conserved element in all GTP binding domains. A tightly conserved Thr residue in Switch I is thought to play a critical role in coordinating the Mg2+ ion required for GTP binding and hydrolysis. Mutation of the Thr residue in eIF5B blocks the activity of the protein in vivo and likewise eliminates the GTPase and in vitro translation activity of the factor. We are currently characterizing several intragenic suppressor mutations that enable the eIF5B Switch I mutant to function in vivo. To assess the role of the long a helix that forms the stem of the eIF5B chalice structure, we have changed the length of the helix by adding or deleting residues. While such mutations did not significantly affect eIF5B activity in vivo, alterations of conserved residues in domain IV drastically impaired cell growth rate. The results suggest that the integrity of domain IV is important for eIF5B function; however, the spacing between domain IV and the N-terminal (cup) domains of the factor is not critical.

Interestingly, the binding site between eIF5B and the translation initiation factor eIF1A maps to domain IV of eIF5B (Choi et al., 2000). The eIF1A is an ortholog of the prokaryotic translation initiation factor IF1, and IF1 has been found to bind to the ribosomal acceptor (A) site. Combining the results of our genetic, biochemical, and structural studies, we can construct a model for eIF5B function during translation initiation. We propose that, following ribosomal scanning to the AUG codon, GTP hydrolysis by eIF2 is coupled to the release of most initiation factors. The eIF1A remains bound in the A site and serves to dock eIF5B on the eukaryotic 40S ribosomal subunit via the interaction between eIF1A and domain IV of eIF5B. In analogy to the translation elongation factors EF-Tu and EF-G, the GTP-binding domain of eIF5B is proposed to contact the large ribosomal subunit. Thus, eIF5B would promote subunit joining by making specific contacts with both the small and large ribosomal subunits. Hydrolysis of GTP by eIF5B and the accompanying conformational changes could reposition the Met-tRNA in the ribosomal P site and trigger release of eIF1A and eIF5B (Dever et al., 2001).

Vaccinia and Pox Virus Pseudosubstrate Inhibitors of eIF2a Kinases in Yeast: Molecular Implications for Kinase-Substrate Interactions
Cao, Dever, Ozatoc
Previous work demonstrated that the vaccinia virus K3L protein and the swine pox C8L protein are pseudosubstrate inhibitors of PKR, the mammalian interferon-induced eIF2 kinase. Expression of either the K3L or C8L protein reduced eIF2a phosphorylation and blocked the toxic effects associated with expression of PKR in yeast. The inhibition of PKR by the K3L and C8L proteins depends on a sequence motif conserved among all K3L homologs and eIF2a. We have extended these studies to mammalian systems by showing that the K3L and C8L proteins inhibit PKR expressed in mammalian cells in a manner dependent on the critical residues required for their anti-PKR functions in yeast (Kawagishi-Kobayashi et al., 2000).

To extend our project, we have isolated K3L mutants that are more potent inhibitors of PKR. Whereas wild-type K3L partially suppresses the toxic effects of PKR in yeast, hyperactive K3L proteins more fully suppress PKR toxicity in yeast. In addition, we have identified 12 PKR point mutants that are resistant to inhibition by K3L and retain the ability to phosphorylate eIF2a. Interestingly, the mutations cluster in the C-terminal lobe of the kinase domain and, based on structural modeling, likely alter points of contact between PKR and its pseudosubstrate (or substrate). We are currently using genetic and biochemical assays to determine how the mutations in PKR and K3L render the kinase either more or less sensitive to pseudosubstrate inhibition.

Substrate Recognition by the eIF2a Kinases
Dey, Liu, Dever
To gain insights into the mechanism of substrate recognition by the eIF2a kinases, we are analyzing GCN2 and PKR deletion mutants of eIF2a to define the minimal substrate for phosphorylation. Residues 1 through 180 of eIF2a appear to be necessary for high-level phosphorylation of Ser-51. Deletion of as few as 10 residues from the amino- or carboxy-terminus of the eIF2a 1 through 180 protein severely impairs phosphorylation of Ser-51 in vivo. The results suggest that the eIF2a kinases recognize a large domain of eIF2a and that kinase-substrate recognition is not limited to a short consensus sequence element flanking the phosphorylation site, as has been demonstrated for other kinases. We used an in vivo screen to identify the residues in yeast eIF2a that are critical for translational regulation and phosphorylation of Ser-51 by the eIF2a kinases. Surprisingly, few substitutions in the residues flanking Ser-51 affect substrate recognition; however, many mutations at these residues alter translational regulation by the eIF2a kinases. In contrast, many mutations at residues 79 through 83, in a sequence element conserved among the K3L homologs and eIF2a, significantly impair phosphorylation of Ser-51. Interestingly, substitution of Ala for Asp-83, 32 residues from the phosphorylation site at Ser-51, completely blocks eIF2a phosphorylation by GCN2 and PKR both in vivo and in vitro. This critical requirement for specific residues in a substrate remote from the site of phosphorylation has not been reported for any other protein kinases. The unprecedented importance of remote sequences, combined with the ease of conducting in vivo analyses of kinase and substrate mutants, makes the yeast eIF2a system ideal for studying protein kinase-substrate recognition.

Heterologous Dimerization Domains Functionally Substitute for the Double-Stranded RNA Binding Domains of the Mammalian eIF2a Kinase PKR
Cao, Elroy-Stein, Dever in collaboration with Ozatoc
The protein kinase PKR is known to dimerize via its double-stranded RNA (dsRNA) binding domains; however, the importance of dimerization for PKR function has been debated. Whereas expression of full-length PKR is toxic in yeast due to phosphorylation of eIF2a Ser-51 and inhibition of translation, expression of the PKR kinase domain alone is not toxic. Interestingly, a GST-PKR kinase domain fusion protein was functional in yeast cells and inhibited translation by phosphorylating eIF2a. Given that the GST protein is known to dimerize, we hypothesized that dimerization was required for PKR function. We constructed fusion proteins that contained the heterodimerization domains from the Xenopus proteins LIM and LDB fused to the PKR kinase domain. When expressed alone in yeast, both LIM-PKR and LDB-PKR were inactive; however, when the fusion proteins were coexpressed in the same cell, they inhibited translation. Thus, PKR activity in yeast depends on dimerization. To extend these studies, we recognized that the drug coumermycin mediates dimerization of GyrB and expressed the GyrB-PKR fusion protein in NIH3T3 cells along with a luciferase reporter construct, which was used to monitor translation. Treatment of the cells with coumermycin elicited a dose-dependent decrease in luciferase activity that correlated with increased phosphorylation of eIF2a Ser-51. The results demonstrated the critical importance of dimerization for PKR activity and suggest that a primary function of dsRNA binding to native PKR is to promote dimerization and activation of the kinase (Ung et al., 2001).