Sequence analysis of Tra1 and TRRAP.
A major obstacle that has impeded structural and functional studies of Tra1 and its human homolog TRRAP is their large size (~430 kDa). The predicted domain structure of Tra1 is well conserved and is similar to those of other members of the PIKK protein family (Fig. ) (1
). PIKK family members consist of five distinct domains, including a C-terminal PI3K catalytic domain that is inactive in Tra1/TRRAP. Four different α-helical domains flank the PI3K domain. Immediately adjacent to the PI3K domain is a four-helix bundle called the FRB domain (FKBP12 rapamycin binding) and an extended helical region called the FAT domain (FRAP, ATM, and TRRAP) (12
). The FAT domain is predicted to form both HEAT (Huntingtin, elongation factor 3, PR65/A, and TOR) and TPR (tetratricopeptide) repeats (4
). The C terminus of PIKK proteins contains a FATC domain (FAT C-terminal) composed of one short and one long helix that are joined by a disulfide bond (20
). However, one or both of these cysteines are missing in Tra1/TRRAP (36
). Finally, the HEAT domain folds into a number of α-helical HEAT repeats that extend from the N terminus to the FAT domain, representing more than half the length of the protein.
FIG. 1. Comparison of Tra1 and TRRAP domains. (A) Modular structure of the Tra1/TRRAP protein. A representative three-dimensional protein structure for each of the five domains is depicted below each arrow with the corresponding Protein Data Bank code. (B) Alignment (more ...)
Web-based repeat and domain prediction methods, including Pfam, Prosite, SMART, REP, ARD, TPRpred, and HHrepID (5
), predicted few or no HEAT and TPR repeats in Tra1/TRRAP. To identify these repeats, we used a prediction strategy recently developed to detect divergent HEAT and TPR motifs that are missed by conventional methods (43
). This approach uses profile-HMM (hidden Markov model) comparisons from the HHpred program to find structural relationships between a protein of interest and proteins with known tertiary structures and has been particularly useful for predicting protein repeats in PIKK protein family members (43
). Using the HHpred approach, we detected 67 or 68 protein repeats in Tra1 and TRRAP (Fig. ). For Tra1, we predicted a nearly tandem array of 53 HEAT repeats extending from the N terminus up to the boundary of the FAT domain. A nearly identical configuration of 54 HEAT repeats was predicted for TRRAP, with the exception of one additional HEAT repeat. Repeat predictions in the FAT domain of Tra1/TRRAP suggest that this domain is composed of 14 TPR repeats, consistent with previous reports for other PIKK protein family members (16
Although Tra1 and TRRAP share a modest 27% amino acid sequence identity, their predicted secondary structures are remarkably similar, and repeat analysis revealed that approximately 98% of the predicted helical repeats aligned well between the two proteins when their protein sequences were compared (see Fig. S1 in the supplemental material) (18
). The Tra1 HEAT domain contains a tandem array of repeats that are interrupted by two predicted disordered regions between HEAT repeats 9 and 10 and HEAT repeats 40 and 41. Disordered regions are also predicted in nearly identical positions of TRRAP, the first between HEAT repeats 9 and 10, which corresponds to a repeated proline stretch, and the second between HEAT repeats 41 and 42. The FRB, PI3K, and FATC domains of Tra1 and TRRAP are also conserved at the sequence level and are predicted to have nearly identical tertiary structures. These observations suggest that Tra1 and TRRAP have similar structures and provide a framework to test the importance of these conserved domains.
Regions of Tra1 important for cell viability.
To identify functional domains within Tra1, we constructed a series of 44 internal deletions spanning the entire length of the protein using the above repeat and domain prediction as a guide. Each deletion removed ~100 amino acid residues (Fig. ). Deletions in the HEAT and TPR domains were specifically designed to remove 1 to 3 repeat units. Since Tra1 is essential for viability, we transformed these mutant constructs into a tra1Δ strain carrying the wild-type TRA1 locus on a URA3-marked plasmid, and transformants were plated on 5-fluoroorotic acid (5-FOA). Approximately two-thirds of the TRA1 mutants were inviable, and many of the remaining viable mutants exhibited slow- and/or temperature-sensitive-growth phenotypes (see Table S1 in the supplemental material).
FIG. 2. Summary of TRA1 deletions and single-amino-acid substitutions in the PI3K domain and their roles in SAGA and NuA4 coactivator function. (A) Schematic of the repeat and domain architecture of Tra1. The color scheme and structural features are the same (more ...)
mutants were found in each of the five domains (Fig. , red). The FRB, PI3K, and FATC domains were all essential for viability. A single deletion at the junction of the PI3K C-terminal lobe and FATC domain gave rise to viable yeast, whereas a complete deletion of the FATC domain was inviable, possibly due to poor protein stability, since this deletion resulted in a truncation in the C terminus (see below). This is consistent with a previous study suggesting that the FATC domain is important for integrity of complexes containing members of the PIKK protein family (36
). We also found that Tra1 regions spanning HEAT repeats 8 and 9, 10 to 25, 20 to 35, and 46 to 53 and TPR repeats 1 to 6 and 10 to 14 were essential for viability. The remaining Tra1 HEAT and TPR repeats, as well as the two aforementioned disordered regions between HEAT repeats 9 and 10 and repeats 40 and 41, were not essential for growth.
A subset of the viable mutants exhibited growth sensitivities consistent with defects in transcription activation (Fig. ; see also Table S2 in the supplemental material). We examined the effects of three stress conditions on growth of these mutant strains including growth on galactose, sensitivity to temperature (37°C), or sulfometuron methyl (SMM). Of the 15 viable mutants, 12 were moderately sensitive to growth at 37°C. These mutants contained deletions in either the HEAT or FAT domains or at the junction between the PI3K C-terminal lobe and FATC domain. Cell growth in the presence of SMM, an inhibitor of amino acid biosynthesis that requires the activation of Gcn4-dependent genes, was impaired for 12 of 14 mutants. Nearly all of these strains also displayed temperature-sensitive phenotypes, suggesting that these Tra1 mutants may cause a defect in heat stress-induced transcription. We observed poor growth on galactose for only 2 out of the 14 viable mutants (e.g., Δ24 and Δ32), suggesting that the Gal4 and Gcn4 activators may interact differently with Tra1. Finally, we tested the dominance of the TRA1 deletion mutants when coexpressed with wild-type TRA1 and observed no significant growth phenotypes under the any of the conditions tested (see Fig. S2 and Table S1).
FIG. 3. Growth phenotypes of selected TRA1 deletion mutants. (Left panels) Viable TRA1 mutants from Table S1 in the supplemental material were compared with a wild-type strain for growth on glucose complete (GC) medium lacking tryptophan, isoleucine, and valine (more ...)
We also analyzed the function of the PI3K domain by modeling the structure of this domain and mutating predicted surface residues (Fig. ). A total of 16 single glutamate substitutions were created, 4 in the N-terminal lobe and 12 in the C-terminal lobe of the PI3K domain. Only two of these mutations, R3432E and R3650E, gave rise to inviable yeast, while the remaining TRA1 PI3K mutants were viable and grew at or near wild-type levels (see Table S3 in the supplemental material). The two inviable substitutions lie very close together when mapped onto a 3D homology model of the PI3K domain, indicating that these residues form an important interaction interface in this region (Fig. , red). None of the viable PI3K domain mutants displayed sensitivity to temperature, SMM, or galactose (Table S2). From this analysis, we conclude that the PI3K domain is essential but contains few regions participating in functional protein contacts.
Regions of Tra1 important for coactivator complex association.
Previous biochemical and genetic studies suggest that Tra1 acts as a molecular scaffold for SAGA and NuA4 complexes (3
). To probe this function of Tra1, we tested the panel of Tra1 deletions for association with SAGA and NuA4 subunits. Yeast strains containing both untagged wild-type TRA1
and Flag-tagged mutant TRA1
were used to test the ability of the viable and inviable Tra1 mutants to associate with other SAGA and NuA4 subunits by coimmunoprecipitation assays. Western analysis of immunoprecipitated Flag-Tra1 revealed that all Tra1 mutants were expressed at levels ranging from 25% to 100% of wild-type levels. Three inviable TRA1
deletion mutants gave rise to truncations at or near the site of the deletion (deletions Δ8, Δ39, and Δ42) (Fig. ; Δ8 is much smaller than wild-type Tra1 and is not shown in the figure). Probing the Western blots for factors coprecipitating with Tra1 surprisingly revealed that all of the inviable Tra1 mutants failed to associate with SAGA subunits Ada1, Spt3, Taf12, and Gcn5 or with NuA4 subunits Esa1 and Yaf9 (Fig. ). Ada1 and Taf12 are important for SAGA integrity, while the other SAGA subunits assayed (Gcn5 and Spt3) are peripheral subunits (83
). The Esa1 and Yaf9 subunits are components of separate functional modules of NuA4 (6
). Even though the bulk of Tra1 is composed of multiple repeat units, these repeats are not functionally equivalent, as deletion of only some repeats results in complex instability. It is clear from these results that yeast cell viability is closely linked to the ability of Tra1 to associate with SAGA and NuA4 and that identical regions of Tra1 are required for association with both complexes.
FIG. 4. Effect of TRA1 mutations on SAGA and NuA4 complex integrity. Western analysis of immunopurified Flag-Tra1. Flag-Tra1 complexes were purified from whole-cell extract and analyzed by Western blotting using the following antibodies: Flag, Ada1, Spt3, Gcn5, (more ...) Regions of Tra1 important for transcription activation.
Several of the viable TRA1 mutants form relatively normal levels of SAGA and NuA4 complexes but display strong growth phenotypes, suggesting that they may be defective in coactivator function. To test this, we used quantitative RT-PCR to measure the mRNA levels of three different classes of genes that are activated by activators Rap1, Gcn4, and Gal4 and coactivators SAGA and NuA4. Examining the different activator and coactivator dependencies is necessary to differentiate between activator-specific, coactivator-specific, and general defects in interaction with the transcription machinery subsequent to Tra1 recruitment.
To study defects in Rap1/NuA4-dependent transcription, we measured transcription levels of three ribosomal protein genes, RPS5
, and RPL11B
. Each of these promoters is dependent on NuA4, as shown by their sensitivities to deletions of NuA4 components but not to disruption of SAGA. For example, deletion of NuA4 subunit EAF1
reduced RNA levels by 60 to 70% for each of the ribosomal protein genes examined (Fig. ). In contrast, these mRNA levels remain relatively unchanged when the SAGA core subunit SPT20
is deleted. Our results are consistent with previous findings that transcription of these ribosomal protein genes is dependent on NuA4 but relatively independent of SAGA (58
). Seven of the viable TRA1
mutants showed decreased ribosomal protein transcription levels ranging from 20 to 75% of the wild-type value (deletions Δ2, Δ5, Δ17, Δ23, Δ24, Δ32, and Δ41) (Fig. A to C). Tra1 deletion mutants Δ17 and Δ32 showed the most severe reduction in ribosomal protein gene transcription, evidenced by a more than 75% reduction in mRNA levels.
FIG. 5. Effect of SAGA and NuA4 mutations on transcription. Strains containing either wild type (WT) or the indicated SAGA or NuA4 subunit deletions were grown under conditions that induce Rap1-dependent genes (A), Gcn4-dependent genes (B), or Gal4-dependent (more ...)
FIG. 6. Effect of TRA1 mutations on transcription. Strains containing either wild-type TRA1 or the indicted TRA1 deletions were grown under conditions that induce Rap1-dependent genes (A to C), Gcn4-dependent genes (D to F), or Gal4-dependent genes (G to I). (more ...)
Failure of some Tra1 mutants to stimulate activated transcription could reflect an impaired step downstream of coactivator recruitment or a defect in interaction with activator. To examine the basis for the transcription defects, we used ChIP to measure the recruitment of polymerase II (Pol II) and the Rap1 activator to the RPL2B promoter. For all seven of the TRA1 deletion mutants tested, Rap1 cross-linking levels were relatively unaffected, and Pol II cross-linking correlated well with transcription levels, suggesting that transcription defects were not the result of reduced activator binding or postinitiation defects (Fig. A and B). These results suggested that either coactivator recruitment or postcoactivator recruitment steps are defective in these TRA1 mutants. To differentiate between these models, we measured the levels of the NuA4 subunits Tra1 and Eaf1 cross-linked to the RPL2B promoter in these strains (Fig. ). Three TRA1 mutants showed decreased cross-linking of Eaf1 and Tra1 at the promoter (Δ2, Δ5, and Δ17), while three other TRA1 mutants showed near-wild-type cross-linking (Δ24, Δ32, and Δ41). Taken together, these results suggest that transcription defects caused by TRA1 deletion mutants Δ2, Δ5, and Δ17 are due to reduced recruitment of NuA4 by activator, possibly due to impaired Tra1 interaction. In contrast, transcription defects in TRA1 mutants Δ24, Δ32, and Δ41 are likely due to postrecruitment defects.
FIG. 7. Effect of TRA1 mutations on recruitment of SAGA and NuA4 to promoters. Strains were grown as described for Fig. , cross-linked, and analyzed by ChIP: activator (A, E, and I), Rpb1 (B, F, and J), Flag-Tra1 (C, G, and K), Eaf1-TAP (D), and (more ...)
We next tested three Gcn4-activated genes: HIS4, ARG1, and ARG4. Each of these genes is SAGA dependent, as revealed by reduced levels of induced transcription in the SPT20 deletion strain (Fig. ). Although transcription of ARG4 is unaffected by mutations in NuA4 subunits, both HIS4 and ARG1 transcription are NuA4 dependent, showing that NuA4 functions in a gene-specific manner at Gcn4-dependent genes. Reduced transcription was observed for six TRA1 mutants (Δ2, Δ5, Δ17, Δ24, Δ32, and Δ41) ranging from 40 to 60% of wild-type levels (Fig. ). Interestingly, the transcription results for the Gcn4- and Rap1-dependent genes were very similar, suggesting that these TRA1 mutations cause similar transcriptional activation defects at both SAGA and NuA4-dependent genes.
To determine the cause of these defects, we used ChIP to measure cross-linking of Gcn4 and Pol II at HIS4. Under activated conditions, Gcn4 cross-linking to the HIS4 promoter was unaltered in any of the TRA1 mutant strains, and Pol II cross-linking correlated well with mRNA levels (Fig. ). Next, we measured the levels of SAGA recruited to the activated HIS4 promoter by ChIP of Tra1 and the SAGA core subunit Spt7. Four TRA1 mutants showed significant defects in SAGA recruitment at HIS4 (deletions Δ2, Δ5, Δ17, and Δ41) (Fig. ). Three of these TRA1 mutants also showed reduced NuA4 recruitment in Rap1-dependent transcription (Δ2, Δ5, and Δ17), suggesting that Rap1 (and/or other activators of ribosomal protein gene transcription) and Gcn4 target overlapping regions of Tra1 to recruit SAGA and NuA4 (Fig. ).
The final group tested included the Gal4- and SAGA-dependent genes GAL1, GAL7, and GAL3. Deletion of SPT20 severely reduced activated mRNA levels of all these Gal4-dependent genes, consistent with several studies that showed that SAGA is an essential coactivator for these genes (Fig. ). In contrast, deletions of NuA4 subunits EAF1 and YNG2 reduced GAL3 transcription by more than 2-fold, while GAL1 and GAL7 transcription remained relatively unchanged. Thus, like the Gcn4-dependent genes, Gal4-activated genes are dependent on SAGA and differentially dependent on NuA4 for their transcription.
The strongest transcription defects among the Gal4-dependent genes were observed for TRA1 mutants Δ24, Δ32, and Δ41, while the remaining Tra1 mutants showed near-wild-type or only slightly reduced mRNA levels (Fig. ). Pol II cross-linking at the GAL1 promoter correlated well with transcription levels for all mutants tested, while Gal4 cross-linking remained relatively unchanged, indicating that transcription defects occur after activation but before transcription initiation (Fig. ). Cross-linking of the SAGA subunits Tra1 and Spt7 at GAL1 remained within 25% of the wild-type level for each of the Tra1 mutants, with the exception of Tra1 mutant Δ24, where cross-linking was reduced by 40% (Fig. ). Thus, Δ24 may disrupt a region of Tra1 that is important for recruitment of SAGA that is unique to the Gal4 activator. Furthermore, these results are consistent with a model where Gal4 and Gcn4 target different regions of Tra1, since the Tra1 mutants defective in recruitment by Gcn4 (Δ2, Δ5, and Δ17) show little or no defect in SAGA recruitment to the GAL1 promoter.
To complement the ChIP assays, we used GST pulldown assays to test whether TRA1 deletion mutations with recruitment defects are defective in activator binding. A GST-tagged fusion protein containing the two Gcn4 activation domains (residues 1 to 134) was mixed with Flag-Tra1, partially purified from whole-cell extracts, and the amount of Flag-Tra1 that coprecipitates with GST-Gcn4 was determined. Compared to binding GST alone, each of the TRA1 deletion mutants copurified with GST-Gcn4 to various degrees (Fig. A and B). TRA1 deletion mutant Δ32 bound to GST-Gcn4 at near-wild-type levels, consistent with little or no defect in SAGA and NuA4 recruitment. The remaining TRA1 deletion mutants reduced activator binding activity, ranging from 50 to 25% of wild-type levels. These results are consistent with the coactivator recruitment defects detected in these mutants, indicating that the reductions in SAGA and NuA4 recruitment in vivo are the direct result of a defect in activator binding.
FIG. 8. TRA1 mutations affect the binding of SAGA and NuA4 complexes to GST-Gcn4 fusion protein. Equivalent amounts of purified Flag-Tra1 coactivator complex were mixed and incubated with glutathione beads bound to GST-Gcn4 1-134 (A) or GST alone (B). Bound fractions (more ...) Regions of Tra1 important for coactivator-mediated histone acetylation.
TRA1 deletions Δ24, Δ32, and Δ41 showed near-wild-type coactivator cross-linking at some promoters but low levels of activated transcription, suggesting a defect downstream of coactivator recruitment. Tra1 deletion mutant Δ32 exhibits the clearest postrecruitment defect that is neither activator nor coactivator specific, as it is observed at RPL2B, HIS4, and GAL1. A common postrecruitment function shared by SAGA and NuA4 is histone acetylation, an important step in the transcription activation process. H3 lysine 9 (H3K9Ac) and H4 lysine 8 (H4K8Ac) acetylation levels, which correlate well with transcription activation, were examined at the RPL2B, HIS4, and GAL1 promoters in five TRA1 deletion mutants: Δ5, Δ17, Δ24, Δ32, and Δ41.
We monitored NuA4-dependent H4 acetylation levels at the RPL2B promoter and SAGA-dependent H3 acetylation levels at the HIS4 and GAL1 promoters. At the RPL2B promoter, H4 acetylation levels were reduced by 50 to 70% in all mutants tested (Fig. , dark gray bars). Similarly, H3 acetylation levels at the HIS4 promoter showed decreased acetylation in each of the TRA1 deletion mutants (Fig. , dark gray bars). In contrast, at the GAL1 promoter, TRA1 mutants Δ5 and Δ17 retained near-wild-type H3 acetylation levels, while TRA1 mutants Δ24, Δ32, and Δ41 were decreased by between 40 and 60%, respectively (Fig. , dark gray bars). These acetylation defects could be caused either by reduced activity of the HAT modules or by destabilization of the association of the HAT modules with SAGA and NuA4. RPL2B, HIS4, and GAL1 transcription are all sensitive to loss of Gcn5, with transcription reduced by 50% or more when GCN5 is deleted (Fig. ). Likewise, RPL2B and HIS4 transcription is reduced by 50% or more in a conditional ESA1 mutant when grown at 37°C (Fig. ).
FIG. 9. Effect of TRA1 deletions on HAT subunit recruitment, histone acetylation levels, and HAT module integrity. (A to C) Strains were grown as described for Fig. and analyzed by ChIP at the indicated genes. Shown are the ChIP results obtained (more ...)
ChIP was used to measure recruitment of the HAT subunits Gcn5 and Esa1 to the RPL2B, HIS4, and GAL1 promoters. For most TRA1 mutants, defects in H3 and H4 acetylation correlated well with defects in HAT recruitment (Fig. ). A striking exception to this trend was TRA1 mutant Δ32, in which Esa1 or Gcn5 cross-linking was near the wild-type level at RPL2B, HIS4, and GAL1, while H3 and H4 acetylation was reduced to 30 to 40% of the wild-type level. This suggests that deletion Δ32, located within the FAT domain, disrupts the HAT activity of both SAGA and NuA4. In contrast, the reduced HAT cross-linking in TRA1 mutants Δ24 and Δ41 does not correlate well with the near-normal Tra1 levels detected at RPL2B and HIS4, suggesting that these Tra1 mutations destabilize association of the HAT modules.
The HAT module of SAGA consists of three subunits: Ada2, Ada3, and Gcn5 (6
). To determine whether SAGA HAT module integrity is disrupted for TRA1
mutants Δ24, Δ32, and Δ41, we immunoprecipitated Flag-Tra1 and assessed the levels of the HAT components that coprecipitate with Flag-Tra1 by Western blotting (Fig. ). TRA1
mutant Δ24 showed clear reductions in Gcn5, Ada3, and Ada2 association (20%, 30%, and 50% of wild-type levels, respectively), while HAT module subunits remained at or near wild-type levels for TRA1
mutants Δ32 and Δ41. To further examine HAT module stability, we immunoprecipitated Gcn5 and quantitated coprecipitation of the various Tra1 derivatives (Fig. ). In this experiment, equivalent amounts of Gcn5 were loaded to more easily compare the amounts of coprecipitating Tra1. A 2-fold reduction was observed in the amount of Tra1 Δ24 coprecipitating with Gcn5 (lanes 2 and 3). In contrast, Tra1 Δ41 showed only a slight decrease in coprecipitating Tra1 (75% of wild-type level) while Tra1 Δ32 showed no defect in Tra1-Gcn5 association. As expected, Ada2 and Ada3 associated normally with Gcn5 in all strains tested, as these three proteins interact independently of Tra1.
Finally, we tested HAT activity of the Flag-Tra1 purified complexes. Immunoprecipitates of Tra1 from wild-type or mutant strains were assayed for acetylation of recombinant H3 at residue K9 as a measure of intrinsic HAT activity (Fig. ). These assays used Flag-Tra1 immunoprecipitates similar to those in Fig. and were normalized to utilize the same amounts of Flag-Tra1 in each HAT assay. As expected, we detected less H3K9 acetylation with TRA1 mutant Δ24, consistent with a HAT module association defect (Fig. , lane 3). However, HAT assays using Flag-Tra1 Δ32 and Δ41 complexes showed activity similar to that of wild-type Flag-Tra1 complexes. Our combined results show that TRA1 mutant Δ24 affects association with the Gcn5 HAT module while the phenotypes of TRA1 mutant Δ32 are clearly caused by a postrecruitment defect.