The Section on Nutrient Control of Gene Expression is studying how the yeast Saccharomyces cerevisiae regulates the expression of its genome according to nutrient availability and other stresses. In response to starvation for any amino acid, purine, or glucose, a regulatory mechanism known as general amino acid control increases the transcription of about one-fifth of the entire genome, including most of the genes encoding amino acid and purine biosynthetic enzymes. Synthesis of GCN4, the transcriptional activator in this system, is stimulated under starvation conditions by a translational control mechanism involving short open reading frames (uORFs) in the GCN4 mRNA leader as well as general translation initiation factors (eIF2 and eIF2B), which stimulate binding of initiator tRNA (tRNAiMet) to the ribosome. The eIF2 forms a stable ternary complex (TC) with tRNAiMet and GTP and binds to the 40S ribosomal subunit, whereas eIF2B is the guanine nucleotide exchange factor (GEF) for eIF2 that recycles inactive eIF2-GDP to active eIF2-GTP after each round of translation initiation. The induction of GCN4 translation depends on phosphorylation of the a-subunit of eIF2 on serine-51 by the protein kinase GCN2, converting eIF2 from a substrate to inhibitor of eIF2B. The phosphorylation of eIF2 by GCN2 reduces the rate of general protein synthesis and simultaneously induces translation of GCN4 mRNA. In this way, the cell can decrease its demand for amino acids while increasing its capacity for de novo production of amino acids. Using GCN4 expression as an in vivo indicator of eIF2 and eIF2B functions as well as genetic and biochemical assays for protein-protein interactions, we are probing the biochemical roles of different subunits of these factors in the recycling of eIF2 and the inhibition of eIF2B by phosphorylated eIF2. We also are analyzing interactions between eIF2 and other initiation factors involved in tRNAiMet binding to the 40S ribosome and recognition of the AUG start codon (the key steps involved in GCN4 translational control), including eIF3, eIF1A, eIF1, and eIF5. With the protein kinase activity of GCN2 highly regulated by amino acid levels, we are attempting to understand the basis for such control by studying interactions between the kinase domain and flanking regulatory regions in GCN2 and probing the functions of its positive effectors GCN1 and GCN20.

To identify all coactivators required for wild-type transcription of GCN4 target genes and to determine which are recruited to the promoter by GCN4, we are also exploring the molecular mechanism of transcriptional activation by GCN4. We wish to identify the particular subunits of each coactivator complex that contact GCN4 directly and those that interact with general transcription factors. Finally, we have conducted a genome-wide analysis of the complete set of genes that are transcriptionally induced by GCN4.

Mechanism and Regulation of Guanine Nucleotide Exchange on Translation Initiation Factor 2 (eIF2) by eIF2B
Krishnamoorthy, Zhang, Hinnebusch in collaboration with Devera and Pavittb
Previously, we showed that the five subunits of eIF2B can be divided into a catalytic subcomplex containing GCD1 and GCD6 and a regulatory subcomplex containing GCD2, GCD7, and GCN3. The GCD1/GCD6 catalytic subcomplex can catalyze efficient nucleotide exchange but is insensitive to the inhibitory effects of phosphorylating the substrate, eIF2-GDP. The GCD2/GCD7/GCN3 regulatory subcomplex has no catalytic activity but can bind to eIF2 and mediate inhibition of the catalytic subcomplex when eIF2-GDP is phosphorylated. Regulatory mutations were obtained in multiple residues of GCD2, GCD7, and GCN3, and in the a-subunit of eIF2, that abrogate the inhibitory effect of eIF2(aP)-GDP on eIF2B function.

To understand the mechanism of eIF2B inhibition by phosphorylated eIF2, we set out to identify the binding pocket in eIF2B for the phosphorylated Ser-51 residue in eIF2a. We found previously that a recombinant form of eIF2a can bind to purified eIF2B if Ser-51 is phosphorylated in vitro. Moreover, we demonstrated that phosphorylated eIF2a (P-eIF2a) binds to the regulatory subcomplex but not to the catalytic subcomplex of eIF2B, as predicted. All three eIF2B regulatory subunits were required for a stable interaction with P-eIF2a. Genetic experiments provided evidence that the a subunit of eIF2 can bind alone to the eIF2B regulatory subcomplex in vivo. Regulatory mutations in residues surrounding Ser-51 in eIF2a, and others in GCD7, all weakened the binding of P-eIF2a to the eIF2B regulatory subcomplex in vitro. The GCD7 mutations also weakened the association of eIF2B with the eIF2 holocomplex. The data support a model in which tight binding of the phosphorylated a subunit of eIF2 to the eIF2B regulatory subcomplex prevents productive association of the b and g subunits of eIF2 with the eIF2B catalytic subcomplex as a means of impeding nucleotide exchange (Krishnamoorthy et al., 2001).

A Subcomplex of the Three Largest Subunits of Translation Initiation Factor 3 Interacts with eIF1 and eIF5 and Promotes Met-tRNAiMet and mRNA Binding to 40S Ribosomes
Nielsen, Phan, Valasek, Hinnebusch
Mammalian eIF3 is a 10-subunit factor that binds to 40S ribosomal subunits in the first step of translation initiation; it also stimulates recruitment of tRNAiMet and mRNA to the ribosome. We showed previously that affinity-purified yeast translation initiation factor 3 (eIF3) contains only five subunits (TIF32, PRT1, NIP1, TIF35, and TIF34), which are orthologs of human eIF3 subunits a, b, c, g, and i, respectively. The purified five-subunit complex is capable of rescuing tRNAiMet binding to 40S ribosomes in mutant yeast extracts lacking functional PRT1 or NIP1 subunits. We also showed that eIF5, the GTPase-activating protein (GAP) for the TC, copurified in stoichiometric amounts with the eIF3 core complex and that substoichiometric amounts of eIF1 (SUI1 in yeast) copurified with yeast eIF3. By conducting protein interaction assays with recombinant subunits, we produced a subunit interaction map for eIF3 in which PRT1 is the scaffold of the complex, the two smallest subunits (TIF34/TIF35) bind to its extreme C-terminus (and also to one another), and the largest subunit (TIF32) binds to a more N-terminal location on PRT1. The remaining subunit NIP1 binds to TIF32 but not to PRT1. In these binding assays, both eIF5 and eIF1 interacted with the N-terminal portion of the NIP1 subunit of eIF3.

Yeast encodes an ortholog of human eIF3j, which was first identified as a dosage suppressor of the rpg1-1 allele of TIF32 and found to be associated with TIF32 in cell extracts. We discovered that high copy-number HCR1 also suppresses certain Ts- alleles of PRT1, indicating a functional interaction between HCR1 and PRT1. In addition, we showed that a fraction of HCR1 is associated with eIF3 as a substoichiometric component and that HCR1 is stably bound to 43-48S initiation complexes in vivo. Protein interaction studies revealed that HCR1 interacts with PRT1 through the RNA recognition motif (RRM) at the N-terminus of PRT1 and makes multiple contacts with TIF32. Interestingly, TIF32 contains an internal domain related to HCR1 that is part of its binding domain for the PRT1 RRM. Thus, HCR1 and the HCR1-like domain (HLD) in TIF32 both interact with the PRT1 RRM. By purifying a trimeric complex containing HCR1, the HLD of TIF32, and the PRT1 RRM, we showed that these interactions can occur simultaneously. Given that overexpression of HCR1 can suppress certain prt1 and tif32 mutations, it appears that direct binding of HCR1 to the two core subunits promotes one or more essential functions of the eIF3 complex (Valásek et al., 2001).

Consistent with the above scheme for protein-protein interactions, we found that two stable eIF3 subcomplexes could be overexpressed and purified from yeast cells. One contained PRT1, NIP1, and TIF32 (PN2 subcomplex) while the other contained PRT1, TIF34, and TIF35 (P45 subcomplex). A PRT1/TIF32 (P2) subcomplex could also be isolated, but a PRT1/NIP1 (PN) subcomplex was not detected. Only the PN2 subcomplex copurified with HCR1 and eIFs 1 and 5. These findings provide in vivo evidence that TIF2 anchors NIP1 to the rest of the eIF3 complex and that NIP1 provides the physical link between eIF3 and HCR1, eIF1 and erIF5. Binding of eIF5 and eIF1 to 40S subunits in cell extracts was impaired by a mutation in PRT1 that reduced 40S association of eIF3. The results provide strong evidence that recruitment of eIF5 and eIF1 to the 40S ribosome depends on eIF3 (Phan et al., 2001).

Additional support for our protein linkage map came from purifying a mutant complex containing an N-terminally truncated PRT1 subunit (prt1-D100) lacking the RRM. As predicted, the complex contained TIF34 and TIF35 but lacked TIF32, NIP1, HCR1, and eIF5, whose interactions with eIF3 depend on the PRT1 RRM. The prt1-D100 product has a dominant-negative slow-growth (Slg-) phenotype. We showed that the phenotype results from sequestration of TIF34 and TIF35 in defective subcomplexes that cannot bind to 40S ribosomes. Consistently, the prt1-D100 phenotype was suppressed by producing more TIF34 and TIF35 (Valásek et al., 2001).

We tested the ability of a heat-inactivated prt1-1 extract to rescue Met-tRNAiMet and mRNA binding to 40S subunits and translation of luciferase (LUC) reporter mRNA by the aforementioned eIF3 subcomplexes. Surprisingly, the PN2 trimeric subcomplex and, to a lesser extent, the binary P2 subcomplex showed substantial activity in all three assays compared with the P45 subcomplex or PRT1 alone. Thus, the critical functions of eIF3 can be carried out effectively by a subcomplex containing only the three largest core eIF3 subunits. The five-subunit eIF3 complex had a higher specific activity than did the PN2 subcomplex, suggesting that TIF34 and TIF35 augment the functions of the three largest subunits in eIF3. As TIF34 and TIF35 are essential in vivo, they may be crucial for maximal translation of certain mRNAs encoding essential proteins (Phan et al., 2001).

Analysis of a Peripheral eIF3 Subunit Implicated in Transcriptional Control
Asano, Phan, Hinnebusch, Shalev, Valasek in collaboration with Akiyoshi,c Pise-Masison,d Radonovich,d He,e Brady,d Watanabe,c and Yamamotoc
The e-subunit in mammalian eIF3 is encoded by int-6, an integration site of mouse mammary tumor viruses that has been strongly implicated in transcriptional control. The fission yeast S. pombe encodes a protein closely related to mammalian eIF3e/Int-6. We found that S. pombe Int6/eIF3e is a cytoplasmic protein that is physically associated with the five core eIF3 subunits and that it can be detected on 40S ribosomes. While the int6D mutant was viable, it grew slowly and had a reduced polyribosome content. These defects were complemented by expression of human Int-6 protein. Thus, it appears that S. pombe Int6 and human Int-6 are peripheral subunits of eIF3 that enhance the essential function of the complex in translation initiation. Interestingly, haploid int6D cells showed unequal nuclear partitioning, possibly due to a defect in tubulin function, while diploid int6D cells formed abnormal spores. These phenotypes could indicate that Int6 is involved in regulating translation of specific mRNAs or that it could have additional functions in transcriptional control (Akiyoshi et al., 2001).

Unlike plants and fission yeast, budding yeast does not encode a protein closely related to human eIF3e/Int-6; however, the product of the Saccharomyces cerevisiae geneYIL071c (dubbed PCI8) contains the Proteasome-COP9 signalosome-eIF3 (PCI) domain found in human eIF3e/Int-6. We showed that PCI8 and human eIF3e/Int-6 interacted with the yeast eIF3 complex in vivo and in vitro by binding to a discrete segment of PRT1/eIF3b and that human eIF3e/Int-6 interacts with the corresponding domain in human eIF3b. The results suggested that PCI8 could be the functional equivalent of human eIF3e/Int-6 in budding yeast. However, deletion of PCI8 had no effect on cell growth or general translation. Accordingly, we investigated the effects of deleting PCI8 on the total mRNA expression profile by oligonucleotide microarray analysis and found reduced mRNA levels for a subset of heat shock proteins in the pci8D mutant. PCI8 may have dual functions in transcription and translational control of specific mRNAs, as postulated for mammalian Int-6, or it may have altogether lost its activity in protein synthesis during fungal evolution (Shalev et al., 2001).

Multiple Roles for the C-terminal Domain of eIF5 in Initiation Complex Assembly and Stimulating the GTPase Activity of the TC
Asano, Hinnebusch, Nielsen, Phan, Shalev, Valasek in collaboration with Donahuef
We previously showed that the eIF5 C-terminal domain (CTD) contains a conserved bipartite motif that mediates its binding to runs of lysine residues in the N-terminus of the b-subunit of eIF2 and its interaction with the N-terminus of eIF3-NIP1. Moreover, we found that such interactions can occur simultaneously in vitro and showed that eIF5 mediates interaction between the eIF2 and eIF3 complexes in vivo. A multifactor complex (MFC) containing eIFs 1, 2, 3, and 5 and tRNAiMet was detected in cell extracts free of 40S ribosomes; moreover, the MFC depended on the eIF5-CTD for its integrity. A mutation in the eIF5 CTD (tif5-7A) that disrupts the MFC also decreased the rate of translation initiation and cell growth, suggesting that the MFC is an important intermediate in translation initiation. We proposed that the constituent components of the MFC bind to the 40S ribosome as a preformed unit and that eIF5 plays an unexpected role in assembly of the initiation complex, in addition to its conventional role as GAP for the TC.

Given that the eIF5-CTD mediates stable association between eIF3 and TC in the MFC and that mammalian eIF3 can bind directly to 40S subunits, it seemed likely that formation of the MFC would enhance binding of TC to the 40S ribosome. In agreement with this prediction, the tif5-7A mutation in the eIF5-CTD reduced binding of Met-tRNAiMet and mRNA to 40S subunits in vitro. Interestingly, the eIF5-CTD bound simultaneously to the eIF4G subunit of the mRNA cap-binding complex and the NIP1 subunit of eIF3. These latter interactions may enhance the association of eIF4G with eIF3 to promote mRNA binding to the ribosome. In vivo, tif5-7A eliminated eIF5 as a stable component of the preinitiation complex and led to accumulation of 48S complexes containing eIF2; thus, conversion of 48S to 80S complexes is the rate-limiting defect in this mutant. We propose that the eIF5-CTD stimulates binding of Met-tRNAiMet and mRNA to 40S subunits by mediating interactions between eIF3 and eIF2 or eIF4G; however, its most important function may be to anchor eIF5 to other components of the 48S complex in a manner required for efficient AUG recognition and GTP hydrolysis during the scanning phase of translation initiation (Asano et al., 2001).

Regulation of the eIF2a Kinase GCN2 by Uncharged tRNA
Dong, Hinnebusch, Hu, Qiu, Sattlegger in collaboration with Francklyng
GCN2 contains regulatory domains located both N-terminal and C-terminal to the protein kinase (PK) catalytic domain; these domains include a region related to histidyl-tRNA synthetase (HisRS), a C-terminal ribosome-binding and dimerization domain (Cterm), and a degenerate kinase domain. The HisRS-related domain can bind to tRNA in vitro and is believed to mediate the stimulation of GCN2 kinase activity by uncharged tRNA in amino acid-starved cells. GCN2 can be activated by starvation for any amino acid; in fact, we showed earlier that purified GCN2 has consistently similar binding affinities for different deacylated tRNAs but that aminoacylation weakened its interaction with tRNAPhe. In addition to the HisRS domain, the Cterm is required for tRNA binding, and we presented evidence that binding of uncharged (but not charged) tRNA to the HisRS and Cterm weakens the association of these regions with the PK domain of GCN2. Furthermore, an activating mutation (GCN2c-E803V) that weakens PK/Cterm association enhanced tRNA binding by native GCN2. These results led us to propose that tRNA binding stimulates GCN2 kinase activity by eliminating an inhibitory interaction between the HisRS/Cterm and PK domains.

The HisRS region contains two dimerization domains (HisRS-N and HisRS-C) that are required for GCN2 function in vivo but are dispensable for dimerization of full-length GCN2, presumably due to dimerization by the Cterm. Residues corresponding to amino acids at the dimer interface of E. coli HisRS were required for dimerization of recombinant GCN2 HisRS-N and for tRNA binding by full-length GCN2, suggesting that HisRS-N dimerization promotes tRNA binding. HisRS-N also interacted with a portion of the protein kinase (PK) domain. A deletion that impaired the interaction destroyed GCN2 function in vivo without reducing tRNA binding, dimerization, or ribosome-binding by full-length GCN2. Thus, it appears that the HisRS-N/PK interaction is critical for kinase activation. In addition, we found that GCN2c mutations in the Cterm activate GCN2 without dissociating the PK/Cterm interaction. Based on these new findings, we now propose that the HisRS-N and Cterm domain remain engaged with the PK moiety in the activated state and that tRNA binding switches the interaction between PK and HisRS/Cterm domains between inhibitory and stimulatory configurations (Qiu et al., 2001).

Activation of GCN2 in amino acid-starved cells requires GCN1, which binds to the N-terminal domain of GCN2. The interaction with GCN2 was localized to a C-terminal segment of GCN1 (residues 2052 through 2428), requiring residue Arg-2259 in this interval. Substitution of Arg-2259 abolished interaction between the N-terminus of GCN2 and the C-terminal domain of GCN1 in vitro, impaired complex formation between native GCN1 and GCN2 in vivo, and eliminated activation of GCN2 in amino acid-starved cells. The Gcn- phenotype of gcn1-R2259A was suppressed by overexpressing GCN2, consistent with the idea that the mutation reduces the affinity of GCN1 for GCN2 and thereby impairs GCN2 activation. GCN1 binds to ribosomes, and its expression correlates with paromomycin sensitivity in yeast, dependent on the ribosome-binding activity of GCN1. Hence, we propose that GCN1 interacts with GCN2 on the ribosome and promotes activation of GCN2 by uncharged tRNA that enters the decoding site (Sattlegger et al., 2000).

Binding of Double-Stranded RNA to human eIF2a Kinase PKR Is Required for Dimerization and Promotes Critical Autophosphorylation of the Activation Loop
Hinnebusch, Zhang in collaboration with Dever,a Mathews,h Nagamura-Inoue,i Ozato,i Romano,j Tianh
Protein kinase PKR is activated by double-stranded RNA (dsRNA) in virus-infected human cells; it phosphorylates eIF2a to inhibit protein synthesis and limit virus propagation. PKR contains two dsRNA binding motifs (DRBMs I and II) required for its activation by dsRNA. While strong evidence suggests that PKR activation requires dimerization, the role of dsRNA in dimer formation has been controversial. By making alanine substitutions predicted to remove increasing numbers of side-chain contacts between the DRBMs and dsRNA, we found that dimerization of full-length PKR in yeast was impaired by the minimal combinations of mutations that impaired dsRNA binding in vitro. Mutation of Ala-67 to Glu in DRBM-I, reported to abolish dimerization without affecting dsRNA binding, destroyed both activities in our assays. By contrast, deletion of a second dimerization region that overlaps the kinase domain had no effect on PKR dimerization in yeast. Human PKR contains at least 15 autophosphorylation sites, but only T446 and T451 in the activation loop were found to be critically required for PKR kinase activity in yeast cells. Using an antibody specific for phosphorylated T451, we showed that T451 phosphorylation is stimulated by dsRNA binding. Our results provide strong evidence that dsRNA binding is required for dimerization of full-length PKR molecules in vivo, leading to autophosphorylation in the activation loop and stimulation of the eIF2a kinase function of the enzyme (Zhang et al., 2001).

Characterization of the GCN4 Transcriptome
Natarajan, Hinnebusch in collaboration with Marton,k Meyer,k Roberts,k Sladek
We conducted genome-wide expression profiling with cDNA microarrays and found that about 1,000 genes were induced and about 1,000 genes were repressed two-fold or more in response to histidine starvation by 3-aminotriazole (3AT). Profiling of a gcn4D mutant and a strain constitutively producing wild-type GCN4 (GCN4c) showed that the majority of the genes depended on GCN4 for their maximum induction or repression by 3AT. A sizeable fraction of genes retained inducibility by 3AT at a reduced level in gcn4D cells, indicating dual control by GCN4-dependent and GCN4-independent mechanisms. While a large fraction of the induced genes contain GCN4 binding sites, most repressed genes do not, implying that repression is indirect. Moreover, given that little repression occurred in nonstarved GCN4c cells that were overexpressing GCN4, repression seems to depend on a reduction in promoter activity in starved cells combined with high-level GCN4 expression. One possibility is that high-level GCN4 squelches a transcription factor that is also down-regulated in starved cells and that the combination of these effects is responsible for strong gene repression by 3AT (Natarajan et al., 2001).

Except for cysteine and genes involved in biosynthesis of the amino acid precursors isocitrate and a-ketoglutarate, GCN4 induced genes in each amino acid biosynthetic pathway. Unexpectedly, GCN4 also induced vitamin biosynthetic genes, possibly to supply cofactors for amino acid biosynthesis. GCN4 activated peroxisomal genes and mitochondrial transporter proteins, likely reflecting the occurrence of certain amino acid biosynthetic steps in these organelles. Furthermore, GCN4 induced autophagy genes and amino acid transporters, presumably to boost the amino acid supply by nonbiosynthetic routes, and certain aminoacyl-tRNA synthetases, in keeping with the idea that charged tRNAs are major end products of GCN4 activation (Natarajan et al., 2001).

Genes in the aforementioned categories accounted for only about one-third of the genes most highly induced by GCN4 in starved cells (four-fold or more); thus, GCN4 induced many genes without obvious functions in amino acid biosynthesis, including those encoding glycogen- and trehalose-metabolizing enzymes. Interestingly, GCN4 was required for normal glycogen accumulation in amino acid-starved cells. GCN4 also induced purine biosynthetic genes and genes encoding a variety of transcriptional activators, protein kinases, and protein phosphatases. Thus, GCN4 seems to mobilize numerous regulatory circuits in response to amino acid starvation, perhaps reflecting the fact that GCN4 translation is stimulated by starvation or stress conditions besides amino acid deprivation. Such conditions might include limitation for purines or glucose, high salinity, and treatment with rapamycin (inhibitor of the TOR kinases) or the alkylating agent methyl methanesulfonate (MMS). Some of the GCN4 target genes may be required for survival under other conditions, e.g., purine biosynthetic genes in purine-starved cells. Similarly, GCN4 activates a subset of the genes induced by rapamycin. Thus, general amino acid control is much broader than previously imagined both in the size and scope of the GCN4 transcriptome and the range of stimuli that induce GCN4 and its target genes (Natarajan et al., 2001).