The eukaryotic translation initiation pathway produces an 80S ribosome bound to mRNA with methionyl initiator tRNA (Met-tRNAiMet) base-paired to the AUG start codon in the ribosomal P-site (Fig.1).
The Met-tRNAiMet is recruited to the small (40S) ribosomal subunit in a ternary complex (TC) with GTP-bound eIF2, to produce the 43S preinitiation complex (PIC). In vitro, this reaction is stimulated by eIF1, eIF1A, eIF3, and eIF5. The eIF3, eIF5, eIF1 and TC can be isolated in a multifactor complex (MFC) from budding yeast whose formation is thought to promote binding of all constituent factors to 40S subunits in vivo. The 43S PIC interacts with the 5’ end of mRNA in a manner stimulated by factors that bind to the m7G cap of the mRNA (eIFs 4E, -4G, -4A, -4B) or poly(A) tail (PABP), and by eIF3, to produce the 48S PIC. The PIC scans the leader until the anticodon of Met-tRNAiMet base-pairs with an AUG codon. Scanning to the AUG is promoted by eIF1, eIF1A and eIF4G. In the 48S PIC, the GTP bound to eIF2 is partially hydrolyzed to GDP and inorganic phosphate (Pi) in a manner stimulated by GTPase activating protein (GAP) eIF5. The recognition of AUG allows Pi release from eIF2•GDP•Pi to complete GTP hydrolysis by the TC. Met-tRNAiMet is released from eIF2-GDP into the P site, allowing subsequent joining of the 60S subunit catalyzed by eIF5B. The eIF2-GDP is recycled to eIF2-GTP by the guanine nucleotide exchange factor eIF2B for reassembly of the TC (reviewed in 1,2).
Functions of the eIFs have been defined primarily by in vitro analyses of partial reactions of the initiation pathway using purified components from mammals or (more recently) budding yeast. While the yeast factors have critical functions in vivo, it is often unclear whether their in vitro activities correspond to their essential functions in living cells, and little is known about how these functions are carried out at the molecular level. Genetic and biochemical analysis of factors involved in the translational control of GCN4 has provided strong evidence that, in living cells, eIF2 is crucial for recruitment of Met-tRNAiMet to 40S ribosomes (ie. 43S assembly), that eIF1A and certain eIF3 subunits stimulate this reaction, and that the GEF eIF2B is required for optimum TC assembly. Phosphorylation of the a-subunit of eIF2 (eIF2[aP]) by GCN2 in starved cells impairs recycling of eIF2-GDP to eIF2-GTP by eIF2B. While reducing general translation, this specifically induces GCN4 translation by a reinitiation mechanism involving 4 uORFs in the mRNA leader (Fig. 2A).
After translating uORF1, ~50% of the 40S subunits resume scanning downstream in both starved and nonstarved cells. In nonstarved cells, all of the subunits quickly rebind TC, reinitiate at uORF4, and dissociate from the mRNA after translating uORF4. When TC levels are reduced by eIF2(aP), a fraction of ribosomes fails to rebind TC until scanning past uORF4, allowing them to reinitiate at GCN4 instead. Consistent with this, the bypass of uORF4 and induction of GCN4 translation in starved cells is suppressed by overproducing all three subunits of eIF2, or all four essential subunits of eIF2B, both conditions expected to elevate TC levels. Morever, GCN4 translation is constitutively derepressed (Gcd- phenotype) in mutants with defects in eIF2 or eIF2B subunits, or Met-tRNAiMet biogenesis, in which TC assembly is impaired. More recently, we showed that truncating the eIF1A C-terminal tail (DC) or overproducing the N-terminal tail (NTT) of eIF3c/NIP1 (c/NIP1) also produce Gcd- phenotypes that are diminished by overproducing all of the components of TC from a high-copy plasmid (hc-TC), suggesting that these mutations delay TC loading on 40S subunits scanning downstream from uORF1, allowing a fraction to bypass uORFs 2-4 and reinitiate at GCN4 in the absence of eIF2(aP) (Fig. 2B) 1.
The molecular mechanism of ribosomal scanning and accurate AUG selection is also amenable to genetic and biochemical analyses using yeast. Genetic studies established that tRNAiMet, the subunits of eIF2, eIF1 and eIF5 all contribute to stringent selection of AUG as start codon during scanning, as mutations in these factors confer a Sui- (suppressor of initiation codon) phenotype signifying increased initiation at UUG start codons. This is recognized by suppression of the histidine requirement conferred by a mutation (the his4-303)in the start codon of the histidine biosynthetic gene HIS4 resulting from aberrant initiation at an in-frame UUG codon (Fig.5).
Previous biochemical analysis of Sui- mutations in eIF2 subunits and eIF5 by Donahue and colleagues suggested that the rate of eIF5-catalyzed GTP hydrolysis by TC and dissociation of Met-tRNAiMet from eIF2-GDP are key determinants of stringent AUG selection 3. Recently, we identified Sui- mutations, and also (Ssu-) mutations that suppress UUG initiation, in eIF1A, implicating this factor in scanning and AUG selection. As scanning is fundamental to GCN4 translational control, isolation of mutations impairing induction of GCN4 (Gcn- phenotype) in eIF5, eIF1A and eIF3 has implicated domains/residues of these factors in scanning or AUG recognition as well 1. Recent genetic and biochemical studies suggest that functional interactions between eIF1A and eIF5 in the PIC regulate scanning 4, and also that eIF1 plays a “gatekeeper” function in preventing non-AUG selection 5, by promoting scanning 6 and blocking Pi release from eIF2-GDP-Pi 7 until AUG enters the P-site (Fig. 6).
We have made progress in elucidating the in vivo molecular functions of eIFs in PIC assembly and identifying critical residues in each factor involved in different steps of the process. These advances were aided by a technique we developed for crosslinking eIFs, Met-tRNAiMet and mRNAs to 40S subunits in living cells to preserve native PICs during fractionation by sedimentation through sucrose gradients. Combining this assay with genetic analysis of GCN4 expression and biochemical characterization of eIF interactions in vivo and in vitro, we provided several lines of evidence that assembly of eIFs 1, 2, 5, and 3 into the MFC enhances 43S PIC assembly in vivo. First, we identified mutations in specific residues of the eIF3c/NIP1 NTT—the segment linking c/NIP1 to the eIF5-CTT, eIF1, and (indirectly) eIF2b—that weaken these interactions in vitro, reduce eIF2 and eIF5 occupancy of native PICs, and confer a Gcd- phenotype suppressible by hc-TC 5 (Fig. 3). Second, we showed that depleting subunits of eIF2 or eIF3 using conditional “degron” alleles, substantially decreases 40S occupancies of all MFC components in native PICs, showing that MFC is rate-enhancing (but not essential) for assembly or stability of PICs in vivo 8. Third, we identified point mutations in surface-exposed residues 93-97 of eIF1 that confer a Gcd- phenotype and impair MFC stability and 40S recruitment of MFC components in vivo (Fig. 3).
Analysis of eIF1 mutations was greatly enhanced by our collaboration with Jon Lorsch and his colleagues, who measured rate constants and equilibrium binding constants for specific reactions in PIC assembly. This allowed us to show that the Gcd- mutations 9,12 (in the unstructured NTT) and G107R (in the globular domain) lower the rate of TC loading on 40S subunits without decreasing 40S binding of eIF1 itself 9 (Fig. 3). We made the same finding for Gcd- point mutations (F131,F131), or complete truncation (DC), of the unstructured CTT of eIF1A 10,11 (Fig. 4).
Interestingly, the 131,133 mutation in eIF1A eliminates the ability of eIF1 to promote TC loading in the reconstituted system, indicating a functional interaction between eIF1 and the eIF1A CTT. A Gcd- point mutation in the OB-fold of eIF1A (66-70) was found to reduce TC loading indirectly by decreasing 40S binding of eIF1A itself, supporting the idea that eIF1A binds in the 40S A-site in a manner similar to that of related, bacterial translation factor IF1 12 (Fig. 4). The Gcd- eIF1A mutations reduce the levels of MFC components in native PICs, further demonstrating the interdependency in eIF binding to 40S subunits in vivo 10,11.
We previously showed that recruitment of eIF3 to 40S subunits in vivo is promoted by N- and C-terminal segments of a/TIF32 and c/NIP1 13. Recently, we implicated the RRM domain in the N-terminus of b/PRT1 in eIF3 recruitment to 40S subunits. Mutating the RNP1 motif (rnp1) weakens PRT1 binding to a/TIF32 and j/HCR1 and lowers 40S binding of all eIF3 subunits except j/HCR1. Overexpressing HCR1 suppresses the Ts- phenotype and 40S binding defect of other eIF3 subunits in rnp1 cells, while deleting HCR1weakens 40S occupancy of eIF3 in vivo. These results show that the j/HCR1 subunit of eIF3 bridges the 40S subunit and the RRM domain in b/PRT1 to enhance eIF3 binding to 40S subunits in vivo 14.
We uncovered a novel function of the soluble ATP-binding cassette (ABC) protein RLI1 in PIC assembly, which is conserved in mammals. This essential protein is closely related to the GCN2 stimulatory factor GCN20 and translation elongation factor eEF3/YEF3. We found that RLI1 associates with eIF3, eIF5 and 40S subunits in cell extracts, and that its depletion from cells decreases translation initiation and lowers 40S occupancy of the MFC without reducing MFC integrity. Mutating the ATP binding sites in the ABCs is lethal and overexpressing one such mutant impairs translation in vivo and in cell extracts. We proposed that ATP-driven dimerization of the two ABCs could promote a conformational change in the MFC that stimulates its 40S binding 15. In collaboration with Michael Dean’s lab (NCI, NIH), we showed that the mammalian homolog of RLI1, ABCE1, interacts with eIFs 2 and 5 in vivo and that its depletion by siRNA in cells, or immunodepletion from extracts, impairs translation initiation. Consistent with this, elimination of the Xenopus homologue in embryos with antisense oligonucleotides arrests development at the gastrula stage 16.
Intrigued by the fact that RLI1, eEF3/YEF3 and GCN20 interact with ribosomes and regulate translation, we investigated the functions of the remaining soluble ABC proteins in yeast, ARB1 and NEW1. We demonstrated that ARB1 shuttles between nucleus and cytoplasm, interacts with pre-ribosomes and ribosome biogenesis factors (in both 60S and 40S branches of the pathway), and that its depletion from cells reduces the rate of pre-rRNA processing and 40S subunit levels. Deletion of NEW1 also alters 40S:60S subunit ratios 17. With our findings on RLI1, ARB1 and NEW1, it is now clear that all members of this specialized subgroup of ABC proteins are physically and functionally linked to ribosomes.
We have also initiated studies on the mechanism of 43S PIC recruitment to mRNA. Using a degron allele to deplete eIF4G1 in a strain lacking eIF4GII, we found that eliminating eIF4G greatly impairs translation initiation, as expected, but does not reduce steady-state recruitment to 40S subunits of two native mRNAs in vivo, RPL41A and MFA2. By contrast, mRNA recruitment was impaired by depleting eIF2 or eIF3 subunits. These results are surprising because it is generally assumed that eIF4G is critical for mRNA recruitment by providing a protein bridge between eIF3 (on the 40S subunit) and the factors bound to the mRNA cap (eIF4E) and poly(A) tail (PABP). Our results imply that this bridge is not essential (although could be rate-enhancing) for recruitment of at least some native mRNAs, and also that eIF2 and eIF3 can stimulate mRNA recruitment by an eIF4G-independent pathway. Depletion of eIF5 led to accumulation of these and other mRNAs in 40S PICs, providing the first in vivo evidence that its GAP function for eIF2-GTP is rate-limiting for subunit joining in vivo 8.
Previously, we uncovered the GCD10/TRM6 and GCD14/TRM61 genes as factors required for repression of GCN4 translation in nonstarvation conditions, and discovered that they comprise the m1A58 tRNA methyltransferase. Lack of this modification in gcd10 or gcd14 mutants reduces tRNAiMet biogenesis and TC formation with attendant derepression of GCN4 18. Work begun in my laboratory by James Anderson and Anna Krueger on suppressors of gcd10 mutations culminated (in Anderson’s lab) in discovery of a nuclear surveillance mechanism that recognizes the hypomodified tRNAiMet, polyadenylates it by TRF4, and targets it for degradation by the nuclear exosome 19. This was the first report describing the nuclear surveillance complex, now called TRAMP, that targets defective transcripts produced by all three RNA Polymerases.
Our analyses of the prt1-1 mutation in b/PRT1 and of mutations in eIF1A have implicated specific residues in these factors that appear to promote scanning during reinitiation on GCN4 mRNA. The prt1-1 mutation in eIF3b leads to accumulation of bulk 48S PICs, signifying a block in the pathway between binding of 43S PICs to mRNA and 60S subunit joining. Assaying different GCN4-lacZ reporters indicates that prt1-1 eliminates the ability of 40S ribosomes scanning downstream from uORF1 to bypass uORF4 and reinitiate at GCN4 when TC levels are reduced by eIF2a phosphorylation—producing a strong Gcn- phenotype. The simplest explanation for these results is that prt1-1 reduces the rate of scanning, increasing the time required to traverse the uORF1-uORF4 interval and thereby compensates for the delay in TC loading produced by eIF2(aP) 20 (Fig. 7A).
In a related development, we identified Ala substitution mutations in the helical domain of eIF1A (residues 98-101) that confer reduced eIF2 association with native PICs and slow-growth (Slg-), which both are rescued by hc-TC. Despite the impaired 40S binding of TC, the 98-101 mutation produces a Gcn- phenotype rather than the expected Gcd- phenotype. We suggested that 98-101 also confers slower scanning between uORFs 1 and 4, which compensates for the delay in TC loading and restores efficient reinitiation at uORF4, with impaired induction of GCN4 (Gcn-). In collaboration Tatyana Pestova’s lab, we obtained evidence supporting this model by determining the effect of 98-101 on scanning in a reconstituted mammalian system in which inhibition of primer extension (toe-printing) is used to map the locations of scanning PICs on specific mRNAs. We found that yeast eIF1A substitutes for the mammalian eIF1A in this assay and that 98-101 reduces the ability of PICs to both migrate from the cap and to bypass an upstream GUG (Fig. 8).
We proposed that aberrant GUG recognition results from a higher dwell time for each triplet in the P-site that increases the chance for non-AUG selection before scanning resumes. In agreement with this, 98-101 increases initiation of a lacZ reporter with a UUG start codon (Sui- phenotype) and decreases leaky scanning of the AUG at GCN4 uORF1 in vivo—both consistent with a higher probability of start codon selection versus continued scanning. Strikingly, the DC truncation and 131,133 point mutations in the eIF1A-CTT likewise increase initiation at UUG and AUGs in vivo, and DC confers defective scanning and elevated GUG selection in the mammalian in vitro system. We concluded that the CTT and helical domain of eIF1A promote scanning, and that a slower rate of scanning in the eIF1A mutants is responsible for both increased UUG initiation and impaired derepression of GCN4 translation 10,11.By measuring the kinetics of eIF1A dissociation from reconstituted 48S PICs, Lorsch and colleagues found that eIF1A binding to the PIC is strengthened (resulting in slower dissociation and less rotational freedom of its CTT), when AUG occupies the P-site and eIF5 is present in the complexes. As both the DC truncation of eIF1A and the SUI5 mutation in eIF5 strengthen eIF1A binding at UUG in this assay, and both mutations increase UUG initiation in vivo (Sui- phenotypes), they proposed that tighter binding of eIF1A characterizes the closed, scanning arrested conformation of the initiation complex at the start codon 4 (Fig. 6a).
In collaboration with Lorsch’s group, we showed that the Sui- eIF1A-CTT mutation 131,133 behaved similarly to DC in strengthening eIF1A binding at UUG, whereas mutations in the unstructured NTT of eIF1A (residues 7-11 or 17-21) have the opposite effect and weaken eIF1A binding with UUG or AUG in the P-site (Fig. 9).
Remarkably, these NTT mutations suppress the Sui- phenotypes of mutations in eIF5 (SUI5)and eIF2b (SUI3-2) and increase leaky scanning of the GCN4 uORF1 AUG (Fig. 10)—both consistent with a reduced probability of start codon selection vs. continued scanning.
Thus, the eIF1A CTT and NTT mutations have opposing effects on start codon selection and binding of eIF1A to the PIC, indicating that the strength of eIF1A association with the PIC is an important determinant of start codon selection in vivo 11.
Several lines of evidence converged recently indicating that eIF1 is a negative regulator of initiation at non-AUG codons. We found that overexpression of eIF1 suppresses the increased UUG initiation in various Sui- mutants 5. Pestova’s group showed that eIF1 blocks PIC assembly at non-AUG triplets in the reconstituted mammalian system 6 and eIF1 also restrains eIF5-stimulated GTP hydrolysis at non-AUGs 21. They mapped the location of eIF1 near the P-site and suggested that it promotes an open conformation of the PIC that is conducive to scanning and restricts base-pairing of Met-tRNAiMet with non-AUG triplets 22. Using the reconstituted yeast system, Lorsch’s group showed that AUG in the P-site evokes a rapid conformational change that increases separation between eIF1 and the eIF1A CTT (detected as loss of FRET between fluorescently tagged forms of eIF1A and eIF1) followed by dissociation of eIF1 from the 40S subunit 23. They also showed that Pi release is stimulated much more than hydrolysis of GTP to GDP-Pi in TC by an AUG in the P-site, that the kinetics of eIF1 dissociation and Pi release are similar, and that the G107R mutation in eIF1 similarly reduces the rates of both reactions 7. This all suggested a model wherein eIF1 blocks non-AUG selection by preventing Pi release, in addition to its other functions in restraining eIF5 GAP function and promoting scanning. All three eIF1 functions would be eliminated when AUG base-pairs with Met-tRNAiMet, as this triggers eIF1 dissociation from the PIC 23.
In collaboration with Lorsch’s and Pestova’s groups, we obtained strong support for the idea that dissociation of eIF1 from the 40S subunit is a key step in start codon selection in vivo. We showed that Sui- eIF1 mutations D83G, Q84P, and 93-97 all decrease eIF1 affinity for 40S subunits and both the Sui- phenotypes and impaired 40S binding of eIF1 are partially corrected by overexpressing the mutant proteins in vivo. Importantly, the 93-97 mutation elevates the rates of both eIF1 dissociation and Pi release from eIF2-GDP-Pi in reconstituted PICs (Fig. 11).
All 3 eIF1 mutations also increase selection of non-AUGs in the reconstituted mammalian system independent of GTP hydrolysis. Remarkably, the eIF1A NTT mutation FL-17-21 that suppresses UUG initiation in Sui- mutants (Ssu- phenotype) decreases (rather than increases) the rate of eIF1 release from reconstituted initiation complexes (Fig. 11). These results indicate that release of eIF1 from the 40S subunit is a critical step in AUG selection in vivo, allowing Pi release and the transition to a closed, scanning-arrested conformation of the PIC, that is modulated by eIF1A 9. We have also identified Sui- and Ssu- mutations in the eIF3c/NIP1 NTT (Fig. 3), that weaken interactions of this eIF3 domain with eIF1 and eIF5, suggesting that eIF3 helps to coordinate the opposing functions of eIF1 and eIF5 in AUG selection 5.
Finally, in collaboration with Mercedes Tamame, we characterized a Gcd- mutation in residue G76 of the 60S subunit protein RPL33A that impairs 40S-60S subunit joining in vivo. The G76R mutation leads to leaky scanning of the inhibitory uORF4 in GCN4 mRNA, presumably by reassembled PICs containing the TC, to elicit derepression of GCN4 translation (Gcd- phenotype). Interestingly, overexpressing tRNAiMet suppresses the Gcd- phenotype of rpl33a-G76R, suggesting that Met-tRNAiMet dissociates from 48S PICs stalled by the delay in subunit joining at the uORF4 start codon, and that this abortive event can be reversed by mass-action to prevent leaky scanning of the uORF 24.
GCN2 exists in a latent form in nonstarved cells and is activated by binding of uncharged tRNAs to the HisRS domain, which in turn interacts with the protein kinase (PK) domain to mediate kinase activation in starved cells. We previously identified dominant-activating GCN2c mutations in residues in or near the hinge segment linking the PK N- and C-lobes (R794G, F842L, E803V) that confer high-level kinase activity independent of the HisRS (tRNA binding) and other flanking domains in GCN2 25. We collaborated with Stephen Burley’s group to determine the crystal structures of wild-type (WT) and R794GPK domains bound to ATP. These structures suggest that the GCN2c mutations eliminate interactions that (in WT) rigidify the hinge and likely impede both interdomain motion and ATP entry to the active site (Fig. 12).
The PK structures represent inactive conformations, however, lacking phosphorylated T882 in the activation loop and displaying improper orientation of the C-helix in the N-lobe 26. Interestingly, the antiparallel mode of PK dimerization in the GCN2 crystals differs dramatically from the parallel mode observed in co-crystals of an activated dimer of human eIF2a kinase PKR bound to substrate eIF2a and phosphorylated in the activation loop 27. Recent work by Dever et al. indicates that dimerization in the parallel orientation is likely required for GCN2 activity 28. Thus, although GCN2 dimerizes constitutively through a remote C-terminal region, activation by tRNA may trigger a shift in PK domain dimerization from antiparallel to parallel mode in order to stimulate autophosphorylation or re-orient the C-helix for kinase activation.
GCN2 activation requires interaction of its N-terminal domain with the GCN1/GCN20 regulatory complex, through a segment in the C-terminal portion of GCN1. Both GCN2 and GCN1/GCN20 associate with elongating ribosomes (polysomes) in cell extracts and the ribosome-binding activity of GCN2 appears to be critical for its kinase function in vivo. GCN1 contains a central domain with similarity to a segment of eEF3, the elongation factor that promotes release of deacylated tRNAs from the ribosomal E site. GCN20 shows strong similarity to the C-terminal portion of EF3, encompassing the two ABCs. Hence, we proposed that GCN1/GCN20 functions on translating ribosomes to facilitate GCN2 activation by uncharged tRNA bound to the ribosomal A site 29. This model is reminiscent of the activation of E. coli RelA protein by uncharged tRNAs on translating ribosomes in the stringent response to amino acid starvation 30. We identified point mutations in GCN1 (M1 & M7, Fig. 13) that specifically impair its association with polysomes and reduce GCN2 activation in starved cells, showing that GCN1/GCN20 act on translating ribosomes to stimulate GCN2.
We also showed that YIH1 protein, related in sequence to the GCN2 NTT, competes with the latter for binding to GCN1 when YIH1 is overexpressed, and thereby impedes GCN1-dependent activation of GCN2 in starved cells (Fig. 13). YIH1 deletion does not activate GCN2, however, ruling out a general function for YIH1 as negative regulator of GCN2. However, YIH1 copurifies with monomeric actin and a genetic reduction in actin impairs GAAC in a manner partly dependent on YIH1. We proposed that dissociation of YIH1 from actin unleashes its GCN2-inhibitory function, and that this might occur specifically at growing bud tips to lower GCN2 activity and stimulate translation initiation at sites of polarized cell growth. Collaborative work with B. Castilho’s group in Sao Paulo indicated that the mammalian protein IMPACT is the functional counterpart of yeast YIH1, and that IMPACT might negatively regulate mammalian GCN2 in certain neuronal cells of the brain 31.
Transcriptional activators stimulate assembly of preinitiation complexes (PIC) at their target promoters by removing repressive chromatin structures and recruiting general transcription factors (GTFs) and RNA Polymerase II (Pol II). Activators carry out these functions indirectly by recruiting coactivators. One class of coactivators is the ATP-dependent nucleosome remodeling complexes, including SWI/SNF and RSC, that expose (or obscure) protein binding sites in promoter DNA. Another class is the histone acetyltransferases (HATs), such as SAGA and NuA4. Histone acetylaton destabilizes chromatin structure and also stimulates recruitment of other coactivators harboring bromodomains. Similarly, it is thought that histone methyltransferases enhance recruitment of coactivators containing chromodomains (CHD) or plant homeodomain (PHD) fingers. A third group of coactivators serve as adaptors to help recruit TATA binding protein (TBP) or Pol II itself, a function generally ascribed to SAGA (for TBP) and the Mediator complex (for Pol II). Mediator further stimulates phosphorylation of Ser5 in the heptad repeats of the C-terminal domain (CTD) of the largest subunit of Pol II (RPB1) by the CDK KIN28 in TFIIH. Transcriptional activation also leads to increased association of certain cofactors with the coding sequences, including the Paf1 complex (Paf1C), which interacts with Pol II and promotes recruitment of histone methyltransferases that target histone H3 on Lys4 (by SET1 complex) and Lys36 (by SET2). Paf1C also promotes Ser2 phosphorylation of the RPB1 CTD in elongating Pol II and thereby stimulates polyadenylation and transcription termination 32.
We are studying the mechanism of transcriptional activation of amino acid biosynthetic genes by GCN4. We showed previously that the activation domain (AD) of GCN4 contains 7 hydrophobic clusters that make additive contributions to transcriptional activation in vivo 33 and stimulate GCN4 binding to SAGA, SWI/SNF, RSC, Mediator and CCR4/NOT complexes in cell extracts 34,36. We also obtained Gcn- mutations impairing activation by GCN4 in one or more subunits of all 5 of these cofactors 35,36 and showed that GCN4 recruits them all to its target gene ARG1 in vivo 36. More recently, we showed that mutations in one or more subunits of these cofactors reduce recruitment of TBP and Pol II by GCN4 to the promoters at ARG1, ARG4 and SNZ1, implicating all five complexes in stimulating PIC assembly. Interestingly, deletion of certain SAGA subunits has a greater impact on recruitment of Pol II versus TBP. Thus, even though TBP binding to the TATA element is required for Pol II recruitment, (which we demonstrated by analyzing a TATA element deletion at ARG1), it appears that SAGA also promotes Pol II recruitment independent of stimulating TBP recruitment 37.
In addition to reducing TBP and Pol II recruitment, the arg1-TATAD mutation lowers recruitment of other GTFs (TFIIB, -IIA, -IIE, -IIF) but has no effect on recruitment of SAGA, Mediator or SWI/SNF to the UAS 38. Thus recruitment of these coactivators by GCN4 is independent of PIC assembly. Consistent with this, our kinetic ChIP analysis showed that, on induction of GCN4 by amino acid starvation, recruitment of cofactors to the UAS precedes TBP and Pol II recruitment to the ARG1 promoter. Despite nearly simultaneous recruitment of SWI/SNF, Mediator, and SAGA to the UAS, we observed strong interdependency in their recruitment by GCN4. Thus, SWI/SNF recruitment is stimulated by SAGA (HAT and non-HAT functions) and Mediator, and recruitment of SAGA is promoted by Mediator and RSC. Recruitment of Mediator is dependent on SAGA at ARG4 and SNZ1 but not at ARG1 38,39 (Fig. 14A).
This extensive interdependency distinguishes GCN4 from the activator GAL4, which recruits SAGA and Mediator independently 40 and requires PIC assembly for SWI/SNF recruitment 41, and also from activator SWI5 that recruits SWI/SNF independently of Mediator and SAGA and requires SWI/SNF for SAGA and Mediator recruitment (at least in late mitosis) 42. Thus, yeast activators exhibit distinct patterns of cofactor interdependency.
Our kinetic analyses of PIC formation in coactivator mutants confirmed that TBP recruitment per se is not sufficient for wild-type promoter occupancy by Pol II and suggested that all four coactivators enhance Pol II recruitment downstream of TBP binding to the promoter. We further uncovered functions for SWI/SNF and SAGA in transcription elongation as mutations in these cofactors had greater effects on Pol II occupancy of coding sequences versus the promoter 39 (Fig. 14B).
Together, these results provide a detailed picture of the GCN4 activation mechanism which differs significantly from those described for other activators, and they extend the range of known functions stimulated by these cofactors in vivo.
Recently, we found that SAGA is associated at high levels with the coding sequences of GCN4 target genes, and also with GAL1, during induction, and that SAGA association with the ORF requires both transcription and Ser5-CTD phosphorylation by KIN28. We further showed that GCN5, most likely in SAGA, functions in transcribed coding sequences to (i) enhance nucleosome eviction from the highly transcribed GAL1 gene; (ii) maintain high-level H3 acetylation in nucleosomes reassembled in the wake of elongating Pol II; (iii) promote Pol II processivity to an extent that increases transcriptional output from an ORF of extended (8kb) length; and (iv) stimulates H3-K4 trimethylation. Interestingly, GCN5 also opposes the effects of several histone deacetylase complexes that are likewise recruited by GCN4 to transcribed coding sequences, presumably to maintain the optimum level of H3 acetylation needed to prevent gene silencing (by hypoacetylation) or activation of cryptic promoters (by hyperacetylation) 43 (Fig. 15).
We made progress on the mechanism of Mediator recruitment by demonstrating that the tail subcomplex containing GAL11/MED15, MED2, and PGD1/MED3 is an in vivo target of the GCN4 activation domain. Deleting each of these subunits impairs recruitment to ARG1 of all Mediator subunits tested. A stable tail subcomplex released from Mediator in a sin4D/med16Dmutant can bind to the GCN4 activation domain in vitro. Importantly, the tail, but not head, subunits of Mediator are recruited by GCN4 in sin4D cells and the function of MED2 in promoting TBP recruitment to the promoter is maintained in the sin4D mutant. Hence, GCN4 can recruit the tail domain independently of the rest of Mediator, and the tail may provide an adaptor function for TBP recruitment 44 (Fig. 16).
Although Paf1C stimulates several important co-transcriptional events, the mechanism of Paf1C recruitment was poorly understood. We discovered that the SPT4 subunit of the yeast equivalent of DSIF, Ser5-CTD phosphorylation by KIN28, and the cyclin-dependent kinase BUR1/BUR2 all promote Paf1C recruitment to elongating Pol II. Consistent with this, spt4Dand bur2Dmutations decrease Paf1C-dependent H3-K4 trimethylation. Since SPT4-Pol II association is independent of both Paf1C and CTD-Ser5P, we proposed that SPT4 (and most likely its DSIF partner SPT5) provide a platform for Paf1C recruitment on elongating Pol II 45 (Fig. 17).
We also showed that the nuclear cap binding complex (CBC) is recruited co-transcriptionally by the m7G cap on nascent transcripts and plays a direct role in preventing polyadenylation at weak termination sites. Similar to NPL3 with which it interacts, CBC carries out its antitermination function by impeding recruitment of subunits of cleavage factor (CF) IA at weak poly(A) addition sites 46 (Fig. 18).
ARG1 is repressed by the ArgR/Mcm1 complex in arginine-replete cells and we found, unexpectedly, that GCN4 recruits all four subunits of the arginine repressor to ARG1 under conditions of isoleucine/valine starvation in which ARG1 is induced by GCN4, but not to GCN4 target genes (ARG4 and SNZ1) unregulated by ArgR/Mcm1. We found that MCM1 and ARG80 reside in a soluble complex lacking ARG81/ARG82 in arg81D cells and are recruited to ARG1 in WT cells independently of arginine and ARG81. By contrast, recruitment of ARG81 and ARG82 was stimulated by exogenous arginine. Thus, it appears that GCN4 constitutively recruits an Mcm1-ARG80 heterodimer and that efficient assembly of the complete 4-subunit repressor complex at the promoter occurs only in arginine excess. We proposed that by recruiting an arginine-regulated repressor, GCN4 can precisely modulate its activation function at ARG1 according to arginine availability 47.
All related news