|Home | About | Journals | Submit | Contact Us | Français|
The cellular role of the Ada2 coactivator is currently understood in the context of the SAGA histone acetyltransferase (HAT) complex, where Ada2 increases the HAT activity of Gcn5 and interacts with transcriptional activators. Here we report a new function for Ada2 in promoting transcriptional silencing at telomeres and ribosomal DNA. This silencing function is the first characterized role for Ada2 distinct from its involvement with Gcn5. Ada2 binds telomeric chromatin and the silencing protein Sir2 in vivo. Loss of ADA2 causes the spreading of Sir2 and Sir3 into subtelomeric regions and decreased histone H4 K16 acetylation. This previously uncharacterized boundary activity of Ada2 is functionally similar to, but mechanistically distinct from, that of the MYST family HAT Sas2. Mounting evidence in the literature indicates that boundary activities create chromosomal domains important for regulating gene expression in response to environmental changes. Consistent with this, we show that upon nutritional changes, Ada2 occupancy increases at a subtelomeric region proximal to a SAGA-inducible gene and causes derepression of a silenced telomeric reporter gene. Thus, Ada2, likely in the context of SAGA, is positioned at chromosomal termini to participate in both transcriptional repression and activation in response to nutrient signaling.
Posttranslational modifications of histones are central to chromatin-dependent functions and create the context in which trans-acting factors perform DNA transcription and silencing, replication, repair, and chromosome segregation. Histone-modifying enzymes are often part of multisubunit complexes identified through biochemical fractionation and genetic analyses. The SAGA complex consists of at least 20 subunits that acetylate histones throughout the genome and at specific genes (reviewed in reference 4). Global histone acetylation patterns resulting from genome-wide acetylation (31) likely underlie chromosomal domain structure. Within this context, SAGA is recruited to specific promoters to induce transcription in response to environmental cues including carbon source utilization, amino acid starvation, and osmotic stress (reviewed in reference 75). Moreover, gene-specific repressive roles for SAGA, such as histone acetyltransferase (HAT)-dependent transcriptional repression of ARG1 (54), have also been observed. Gcn5 is the enzymatic activity in SAGA that preferentially acetylates histone H3 at lysines 9 and 14 and histone H2B in vivo (30, 71, 80). Ada2 functions as a coactivator with Gcn5 to augment its HAT activity (reviewed in reference 67).
ADA2 was identified through a screen for mutants that suppress the toxicity of elevated VP16 expression in Saccharomyces cerevisiae (8). Further studies revealed a role for Ada2 in transcriptional activation, in part through its interaction with Gcn5 (41) to increase the HAT activity of Gcn5 (5, 14, 30). A role for Ada2 in activation also comes from studies demonstrating its interaction with transcriptional activators Gal4 and Gcn4 (6) and with TATA-binding protein (TBP) (28). These interactions at SAGA-dependent genes facilitate preinitiation complex formation, perhaps by bridging activators and TBP (6) to induce transcription (reviewed in references 67 and 74).
No role for Ada2 independent of Gcn5 has been characterized in Saccharomyces cerevisiae, although multiple forms of Ada2 exist in Drosophila melanogaster (37, 44), plants (77), mice (37), and humans (7). In Drosophila, Ada2α and Ada2β are required for normal development but have distinct functions (46). Ada2β associates with Gcn5 in a SAGA-like complex and is required for viability and characteristic histone H3 acetylation (23, 51). Ada2α is a subunit of the Gcn5-containing ATAC complex (27) of unknown function, and mutations in Ada2α cause decreased H4 acetylation (17). Whether the metazoan Ada2 homologs have functions independent of Gcn5 remains to be determined.
Here we provide evidence for a novel function of Ada2 in yeast that is distinct from its transcriptional activation role with Gcn5. Ada2 physically associates with transcriptionally repressive heterochromatin at telomeres and the ribosomal DNA (rDNA) in the context of SAGA or a related HAT complex to restrict the inward spread of heterochromatin source. In physiological response to changes in nutrient levels, Ada2 occupancy increases at subtelomeric regions and causes derepression of a normally silenced telomeric gene.
Strains are listed in Supplemental Table 1 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm. Gene deletions were generated by standard chromosomal replacement methods. The strain expressing Ada2-18Myc from its chromosomal locus was provided by M. P. Cosma (18). The strain expressing Gcn5-9Myc from its chromosomal locus was provided by F. Robert (55). The strain carrying the HHT3-hhf2K-16A plasmid was generated by PCR mutagenesis by M. Ruault in the histone deletion strain provided by M. Smith (64). To generate 3XHA-SAS2 (pLP 1872), DNA encoding a three-hemagglutinin (HA) tag was amplified, cut with BglII, and cloned into the 5′ BglII site of SAS2 in pRS316 (CEN URA3). The function of the HA-tagged Sas2 was confirmed by suppression of the MATa sir1Δ sas2Δ mating defect (53), and expression was confirmed by immunoblotting (LPY 9328, LPY 9326). Plasmids pMA424 (a Gal4p DNA binding domain [GBD] control vector expressing amino acids 1 to 147 of the DNA binding domain of Gal4) and GBD-Sir1 (Sir1 fused to the GBD), used for telomere targeting assays, are described in reference 15 and were provided by R. Sternglanz. The plasmid encoding LexA-Ada2 (pLP 1425) was provided by S. Berger.
For telomeric silencing assays, cultures were grown for 72 h and were plated in fivefold serial dilutions onto synthetic complete (SC) medium in order to assess growth or onto SC medium containing 5-fluoroorotic acid (5-FOA) in order to assess levels of URA3 expression (silencing). For ADE2 sectoring assays, strains were grown for 72 h, diluted, and plated onto SC medium containing 10 mg/liter adenine. Plates were incubated first at 30°C for 72 h and then at 4°C for 14 days for optimal color development. For cryptic mating type silencing assays, expression of HML and HMR was assessed by a standard mating assay. For rDNA silencing assays, cells harboring the ADE2-CAN1 rDNA reporter cassette were plated onto selective medium in the absence or presence of 16 μg/ml canavanine and were then incubated at 30°C for 72 h. For alternative carbon source silencing assays, cells were first grown in 2% glucose for 3 days and then plated onto SC medium containing either 2% glucose or 2% galactose. Plates were incubated at 30°C for 3 days.
Chromatin immunoprecipitation (ChIP) assays were performed as described previously (24). Input (INP) and immunoprecipitated (IP) DNAs were analyzed by quantitative PCR (Q-PCR) using the MJ Research Opticon2 system. Oligonucleotides used for Q-PCR are listed in Supplemental Table 2 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm. A standard curve for Ada2p binding was generated from twofold dilutions of INP DNA from Ada2-18Myc cells and untagged cells. Changes in Ada2 occupancy were calculated from the average of data points within the linear range of the assay (plotted on a log scale) from two or more independent experiments, by dividing the tagged Ada2-Myc IP DNA/untagged IP DNA ratio by the Ada2-Myc INP DNA/untagged INP DNA ratio at each location analyzed along telomere VIR. Changes in Sas2 and Gcn5 occupancy were calculated in a similar manner; 3XHA-Sas2 was immunoprecipitated with monoclonal anti-HA (12CA5; Babco), and occupancy was compared to that for an untagged strain containing sas2Δ carrying an HA vector control plasmid. Ada2-18Myc (18) and Gcn5-9Myc (55) were expressed from their endogenous loci and immunoprecipitated with anti-Myc antiserum. The silencing proteins Sir2 and Sir3 were immunoprecipitated with polyclonal anti-Sir2 serum (2916, bleed 8) and anti-Sir3 serum (7996) at 1:2,000. Histones were immunoprecipitated with anti-acetylated H3 (K14) (Upstate catalog no. 07-353) and anti-acetylated H4 (K16) (Upstate catalog no. 06-762) at 1:1,500. For galactose inductions, cells were grown in raffinose for 16 h; then they were pelleted, resuspended in SC medium containing 2% galactose, and incubated for 20 min prior to the addition of formaldehyde (38).
For the coimmunoprecipitation (co-IP) experiments for which results are shown in Fig. Fig.3C,3C, ,22 × 109 cells were lysed with glass beads in buffer A (50 mM HEPES [pH 7.5], 250 mM NaCl, 5 mM MgCl2, 10% glycerol) and protease inhibitors. Lysates were immunoprecipitated with anti-Myc (1:5,000) overnight at 4°C; then they were incubated with protein G-Sepharose (Amersham) and washed four times with buffer A containing 1% glycerol. Beads were pelleted, resuspended in loading buffer, and separated on a 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel. Proteins were transferred to nitrocellulose membranes, blotted with anti-Sir2 (1:5,000), incubated with horseradish peroxidase-conjugated anti-rabbit serum (1:10,000; Promega), and developed by enhanced chemiluminescence (Perkin-Elmer). Reverse co-IPs were performed as described above, except that anti-Sir2 (1:3,000) was used as the primary antiserum for immunoprecipitation, protein A Sepharose CL-4B beads (Amersham) were used to isolate the immune complexes, and anti-Myc (1:8,000) was used for immunoblotting. For the analysis of ChIP samples in Fig. Fig.4D,4D, one-half of the cross-linked chromatin immunoprecipitated with Sir2 (1:1,000) was washed as described above and dissolved in 3× sample buffer. Samples were incubated at 95°C for 30 min to reverse cross-links, subjected to SDS-PAGE, transferred to nitrocellulose membranes, and probed with anti-Sir2 serum (1:5,000). For the protein immunoblotting of whole-cell extracts (see Fig. Fig.4D),4D), 1 × 108 cells were lysed by glass beads in lysis buffer (0.3 M sorbitol, 0.1 M NaCl, 5 mM MgCl2, 10 mM Tris [pH 7.4], and protease inhibitors), resuspended in 1× loading buffer, and separated on a 7.5% SDS-PAGE gel. Proteins were transferred to nitrocellulose membranes, blotted with anti-Sir2 (1:3,000) or in parallel with anti-tubulin (1:4,000) as a loading control, incubated with horseradish peroxidase-conjugated anti-rabbit serum, and developed by enhanced chemiluminescence as described above. For immunoblotting of Ada2 levels (see Fig. Fig.7B),7B), whole-cell extracts of wild-type cells, either untagged or expressing Ada2-Myc from its chromosomal location, were grown in 2% glucose, 2% raffinose, or 2% galactose, lysed, separated by gel electrophoresis as described above, and probed with anti-Myc (1:10,000) or anti-tubulin (1:4,000) as a loading control.
Cellular mRNA was isolated at an optical density of 0.8 from cells grown in 2% glucose or from cells first grown in 2% raffinose and then shifted to 2% galactose for 20 min. RNA was harvested using an RNeasy kit (Qiagen), followed by first-strand cDNA synthesis using the TaqMan reverse transcription kit (Applied Biosystems) with random hexamers as primers. Reaction mixtures lacking reverse transcriptase were included as controls for endogenous DNA that was left undigested by on-column DNase treatment during RNA isolation. First-strand cDNA reaction mixtures were diluted 1:50, and 5 μl was used as the template in Q-PCR. Primers amplifying an internal region of the telomeric open reading frame (ORF) YFR057W and the subtelomeric HXK1 gene on telomere VIR, or the HXK2 gene on telomere VIIL, were used to determine relative steady-state mRNA levels expressed from these ORFs in wild-type and mutant cells. Q-PCR values were normalized by subtracting the values for the control lacking reverse transcriptase and dividing by the corresponding ACT1 value. The value for the wild type grown under glucose conditions was set to 1.0, and the relative RNA levels for each locus were calculated accordingly.
While investigating HAT complex functions in S. cerevisiae, we observed that deletion of ADA2 caused the loss of chromatin-mediated telomeric silencing. Wild-type cells grew well on 5-FOA (Fig. (Fig.1A)1A) due to repression of a URA3 reporter gene on telomere VIIL (26). In contrast, ada2Δ cells were unable to grow on 5-FOA, indicating a loss of silencing. The ada2Δ silencing defect was unexpected, because previous reports showed no defect upon deletion of GCN5 (72), the HAT with which Ada2 functions in the SAGA complex. The gcn5Δ isolate shown in Fig. Fig.1A1A reproduces this observation, although some variability in silencing among gcn5Δ isogenic strains was observed (data not shown).
To determine if the ada2Δ silencing defect was specific for this URA3 reporter gene or for telomere VIIL, silencing of an ADE2 reporter at telomere VR was visualized by a sectoring assay (26). Wild-type colonies sector pink (ade−) and white (ADE+) due to epigenetic switching of transcription states at telomeres (Fig. (Fig.1B).1B). Deletion of ADA2 yielded exclusively white colonies, comparably to a sir2Δ control strain where silencing was completely disrupted, In contrast, deletion of GCN5 yielded pink and white sectored colonies, similar to wild-type colonies. Thus, deletion of ADA2 caused constitutive expression of normally silenced telomeric reporter genes, independent of either the gene or the telomere assayed. This Ada2 telomere function is the first role characterized for Ada2 that is distinct from Gcn5.
To evaluate if other SAGA components were required for telomeric silencing, selected subunits were deleted and assayed for silencing (Fig. (Fig.1C).1C). Deletion of ADA3, a member of the SAGA subgroup (GCN5, ADA2, and ADA3) that mediates Gcn5 HAT activity, caused a silencing defect. Likewise, deletion of SPT7 and ADA1 (data not shown), which maintain SAGA complex integrity, caused a telomeric silencing defect similar to that of control strains with null mutations in sir2 or sir3 (Fig. (Fig.1C).1C). In contrast, deletion of SPT3 or SPT8, whose gene products regulate TBP association with SAGA target genes (21, 40, 68), did not disrupt silencing. Spt8 is distinguished by its selective absence in the SAGA-related complex SLIK/SALSA, which contains a C-terminally truncated form of Spt7 (49, 66, 78).
To further assess the complex in which Ada2 was mediating telomeric silencing, SPT7 mutants that alter the relative cellular levels of SAGA and SLIK/SALSA (78) were assayed for silencing. The spt7-400 mutant was the most severely defective for silencing (see Fig. S1 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm). This mutant, which lacks both the Spt7 C-terminal cleavage site and the motif conferring Spt8 association, is depleted of SAGA in vivo (78), Thus, it appears that the ada2Δ silencing defect is not mediated exclusively through SLIK/SALSA. Together, these observations support a role for a subset of SAGA subunits in silencing, apart from their known transcriptional activation functions.
Because Ada2 binding has been observed at another silenced locus in yeast, the cryptic mating type locus HMR (45), we asked if deletion of ADA2 caused mating defects. An ada2Δ strain was not defective for mating in either a MATa or a MATα background (Fig. (Fig.2A).2A). Furthermore, deletion of ADA2 did not render a MATa sir1Δ strain mating defective (Fig. (Fig.2A),2A), as is characteristic of a sas2Δ mutant also defective for telomeric silencing (53). A more extensive mating analysis of single and double mutant strains is shown in Fig. S2 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm. The observation that a MATa sir1Δ ada2Δ strain did not phenocopy the MATa sir1Δ sas2Δ mating defect indicates that deletion of ADA2 does not simply decrease the expression of Sas2, thereby acting indirectly in telomeric silencing.
To determine if Ada2 participated in silencing at the rDNA array, an ada2Δ strain harboring an ADE2-CAN1 reporter cassette near the 25S gene (22) was assayed. Silencing was assessed by growth on canavanine, which confers toxicity through expression of CAN1 upon disruption of rDNA silencing. Wild-type and gcn5Δ cells grew well in the presence of canavanine (Fig. (Fig.2B),2B), due to effective silencing of this locus. However, ada2Δ cells, like the set1Δ control strain (11), showed impaired growth (Fig. (Fig.2B),2B), indicating a loss of rDNA silencing. Thus, the assays utilized here show that Ada2 functions in silencing at telomeres and the rDNA.
Because Ada2 augments Gcn5 HAT activity yet acts distinctly from Gcn5 in silencing, we considered the possibility that Ada2 was acting in the context of another HAT. Sas2 is the MYST family (named for yeast and human genes MOZ, YBF2 [SAS3], SAS2, and Tip60) H4 K16-specific HAT, with a well-documented role in telomeric silencing (reviewed in reference 76). Deletion of ADA2 or SAS2 caused telomeric silencing defects of similar magnitudes (Fig. (Fig.3A).3A). To determine whether these proteins acted directly at the telomere, ChIP assays were performed on epitope-tagged Ada2 and Sas2. Ada2 was enriched sevenfold relative to levels in an untagged control strain at 200 bp from the telomere (Fig. (Fig.3B),3B), indicating that Ada2 acted directly at telomeres. The binding of Ada2 was maximal at the chromosomal terminus and decreased to fivefold at 1 kb and 5 kb from the telomere.
To determine if Ada2 was in a position to enhance the HAT activity of Sas2, a function analogous to its role with Gcn5, Sas2 occupancy was also assessed by ChIP. Previous studies showed that deletion of SAS2 causes decreased histone H4 K16 acetylation at telomeres (35, 70), suggesting that Sas2 may act directly at telomeres. However, no Sas2 enrichment was detected at points surveyed between 200 bp and 5 kb from the telomere end (Fig. (Fig.3B).3B). Thus, it is unlikely that Ada2 and Sas2 stably interact there, although they may participate in the same or parallel pathways in maintaining telomere structure and function.
Because Ada2 is present at the telomere, it was possible that it interacts with silencing proteins in telomeric heterochromatin. Sir2 is a histone deacetylase that resides at telomeres and appears to counteract the H4 K16 HAT activity of Sas2 to establish heterochromatic boundaries (35, 70). Co-IP experiments demonstrated a small but detectable pool of Ada2 associated with Sir2 (Fig. (Fig.3C).3C). Control strains lacking Sir2 and Ada2-Myc showed no detectable background, although more than 50% of the Ada2 was immunoprecipitated from cellular extracts (Fig. (Fig.3C,3C, bottom). A reciprocal co-IP, in which Sir2 was first precipitated and then probed for Ada2, also showed an interaction of Ada2 with Sir2 (Fig. (Fig.3C).3C). Thus, Ada2 promotes silencing through chromatin association with the silenced loci. Notably, only a fraction of Ada2 associates with Sir2, consistent with the fact that both Ada2 and Sir2 reside in multiple committed complexes at other chromosomal locations. Due to the low cellular abundance of Sir3, an interaction of Ada2 and Sir3 was detectable only when the expression of both these proteins was elevated (data not shown). To further investigate the requirements for Ada2 localization, Ada2 telomeric occupancy was assessed by ChIP in sir2Δ, sir3Δ, and sas2Δ strains. These studies demonstrated that Ada2 telomere association was not dependent on Sir2, Sir3 (see Fig. S3 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm), or Sas2 (data not shown), demonstrating that Ada2 association at this locus was independent of a functional Sir silencing complex.
The Sir2, Sir3, and Sir4 proteins, along with other telomere residents, bind as a hetero-oligomeric complex that creates repressive chromatin and silences embedded genes (reviewed in reference 59). Both in vivo and in vitro data suggest that Sir association not only creates but is sensitive to the acetylation state of the histone H4 N terminus (reviewed in reference 43). Sir2 deacetylates histone H4, which in turn promotes the binding of Sir3, a telomere resident that preferentially binds hypoacetylated histone H4 K16. The spreading of Sir3 and repressive chromatin is postulated to be opposed by Sas2, which acetylates H4 K16 to demarcate a boundary between heterochromatic and euchromatic regions (35, 70). From these data, a titration model in which Sir3 moves inward from the telomere, thereby alleviating its repressive effect at telomere-proximal sequences, was proposed in order to explain the sas2Δ telomeric silencing defect.
To determine whether the ada2Δ telomeric silencing defect occurs by a similar mechanism, Sir2 and Sir3 telomere occupancies were analyzed by ChIP in ada2Δ and sas2Δ cells. Sir2 occupancy in wild-type cells followed a characteristic pattern (Fig. (Fig.4B),4B), as previously published (70). Sir2 was most abundant at the telomere-proximal region (200 bp from the telomere), decreased to 63% of maximal binding at 1 kb and to 20% at 5 kb, and was at background levels 20 kb from the telomere (Fig. (Fig.4B).4B). In an ada2Δ strain, Sir2 occupancy shifted inward from the telomere (Fig. (Fig.4C).4C). The binding of Sir2 was similar to that of the wild type at 200 bp but increased at 1 and 5 kb to 3.5-fold, before decreasing to 2-fold enrichment at 20 kb from the chromosome terminus. A sas2Δ strain exhibited a similar but not identical pattern of Sir2 spreading. Sir2 occupancy was unchanged at 1 kb and reached twofold enrichment relative to the wild type at 5 kb (Fig. (Fig.4C).4C). This pattern of Sir2 redistribution was evident in a gcn5Δ strain, although in this case Sir2 binding at 5 kb was about 50% of that observed in an ada2Δ strain. Control immunoblotting experiments demonstrated that the ada2Δ telomeric silencing was not due to decreased chromatin-associated or soluble Sir2 abundance (Fig. (Fig.4D4D).
Spreading of Sir3 inward from the telomere was also observed in an ada2Δ strain, and the pattern was similar to that of Sir2 redistribution. The level of Sir3 binding in ada2Δ cells was 70% of wild-type binding at 200 bp and increased at 1 kb and 5 kb to 2-fold enrichment before decreasing to 1.2-fold relative to wild-type binding at 20 kb from the chromosomal terminus (Fig. (Fig.4C).4C). These observations demonstrate that inward shifting of the heterochromatin proteins Sir2 and Sir3 occurs in both ada2Δ and sas2Δ cells, supporting the titration model of defective silencing.
The titration model is independently supported by two additional observations. First, an increased gene dosage of SIR3 selectively rescued the telomeric silencing defect of an ada2Δ strain (see Fig. S4A posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm). Second, telomere-proximal targeting of a GBD-Sir1 fusion protein, postulated to increase the local concentrations of Sir proteins (15), rescued ada2Δ and sas2Δ silencing defects in a Sir-dependent manner (see Fig. S4B at the URL mentioned above). The mechanisms by which Ada2 and Sas2 regulate silencing at the telomere are apparently distinct, because an increased Sir3 dosage did not rescue the sas2Δ defect, whereas increased Sir4 expression did, to a modest extent (see Fig. S4A at the URL mentioned above). Additionally, an increased gene dosage of ADA2 did not rescue the sas2Δ telomeric silencing defect, and conversely, an increased gene dosage of SAS2 did not rescue the ada2Δ defect (see Fig. S4A at the URL mentioned above). This lack of cross-complementation underscores the conclusion that although both Sas2 and Ada2 are required for silencing, their roles are distinct.
Whereas our data support the titration model, we note that Sir protein spreading is not sufficient to produce a telomeric silencing defect. First, ada2Δ and sas2Δ strains exhibited only a minimal decrease in Sir protein binding at the chromosomal terminus where the Sir repressive effect is exerted. Second, both ada2Δ and gcn5Δ cells showed maximal spreading of the Sir proteins to the 5-kb region, but only loss of Ada2p caused a telomeric silencing defect.
Because the acetylation state of histone H4 K16 is a critical determinant of telomeric silencing (reviewed in reference 43), acetylation of this residue was analyzed in ada2Δ cells. The level of H4 K16 acetylation 200 bp from telomere VIR was assigned a value of 1.0 and was used to calculate differences as a function of the distance from the end of the telomere. In wild-type cells, H4 K16 was relatively hypoacetylated at 200 bp and 1 kb from the telomere (Fig. (Fig.5A),5A), consistent with Sir2 binding and its associated deacetylase activity. H4 K16 acetylation was elevated 17-fold at 5 kb from the telomere, near IRC7, a nitrogen source-regulated gene (60). This substantial increase may constitute a heterochromatin barrier due to Sas2-mediated H4 K16 acetylation, which remained sevenfold higher than at the telomere at 20 kb. Indeed, Sas2-dependent H4 K16 acetylation is required for efficient subtelomeric incorporation of the H2A.Z variant (62), which has been shown to be important for restricting the spread of telomeric heterochromatin into euchromatic regions (3).
In an ada2Δ strain, H4 K16 acetylation did not change significantly at 200 bp or 1 kb (Fig. (Fig.5A)5A) but decreased to 50% at 5 kb, where wild-type acetylation is highest. In a sas2Δ strain, H4 K16 acetylation decreased to 70% at 200 bp and 1 kb from the telomere (Fig. (Fig.5A).5A). As in ada2Δ mutants, H4 K16 acetylation decreased sharply at 5 kb to 10% of the wild-type level in sas2Δ strains. Thus, H4 K16 acetylation decreased at 5 kb, either by deletion of SAS2, the HAT with substrate specificity for H4 K16, or by deletion of ADA2. This decrease in H4 K16 acetylation likely reflects the inward shifting of the deacetylase Sir2 in ada2Δ and sas2Δ cells (Fig. (Fig.4C).4C). Consistent with this, RT-PCR analysis of steady-state mRNA levels expressed from the subtelomeric genes HXK1 (15 kb from telomere VIR) and HXK2 (25 kb from telomere VIIL) showed that HXK1 RNA levels were decreased 1.5- to 2-fold in ada2Δ and sas2Δ cells, while those of the more internal HXK2 gene were decreased to a lesser extent (see Fig. S5 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm).
The role of N-terminal histone H3 K14 acetylation in telomeric silencing is less clear. Subtelomeric gene repression requires the histone H3 N terminus (42), and histone H3 sites K9, K14, and K18 are hypoacetylated in subtelomeric regions (9, 56). Because histone H3 K14 acetylation is associated with gene activation, we assayed its distribution. In wild-type cells, H3 K14 was hypoacetylated at 200 bp and was assigned a value of 1.0, relative to which H3 K14 acetylation increased to eightfold at 5 kb and 20 kb (Fig. (Fig.5B5B).
In an ada2Δ strain, H3 K14 acetylation levels were similar to wild-type levels at all points surveyed (Fig. (Fig.5B).5B). In contrast, sir3Δ cells showed a dramatic increase in H3 K14 acetylation at 200 bp and 1 kb from the telomere (Fig. (Fig.5B),5B), suggesting that loss of the telomeric SIR complex allows H3 acetylation. Two HATs that are likely to mediate this H3 acetylation are Gcn5 and Sas3. They have similar histone substrate specificity and are synthetically lethal in combination with each other (30). Additionally, they are recruited to similar classes of genes (58). However, neither deletion of GCN5 nor deletion of SAS3 individually in a sir3Δ background abrogated the increase in H3 K14 acetylation (data not shown). Thus, Gcn5 and Sas3 may be functionally redundant here, or another HAT may be responsible for targeting this H3 site (36).
sas2Δ cells also showed increases in telomeric H3 K14 acetylation at 200 bp and 1 kb, although to a lesser extent than sir3Δ cells. This may reflect incomplete inward migration of Sir proteins. However, the effects of loss of H4 K16 acetylation activity by Sas2 were not solely responsible for this change in H3 K14 acetylation levels, since a strain expressing a histone H4 K16A mutant as its sole source of H4 did not exhibit increased H3 K14 acetylation (Fig. (Fig.5B).5B). Thus, telomeric heterochromatin hypoacetylated at H3 K14 is susceptible to significantly increased acetylation at this site upon disruption of the SIR silencing complex. However, H3 K14 hyperacetylation does not necessarily correlate with a silencing defect, because an ada2Δ strain did not show increased terminal H3 K14 acetylation.
Yeast has an intricate nutrient-sensing system that responds to environmental changes at the transcriptional, metabolic, and developmental levels (reviewed in reference 79). SAGA is one of the key HAT complexes, regulating the expression of about 10% of yeast genes in response to changes in the carbon source, nutrient starvation, and environmental stress (reviewed in reference 75). Some of these genes are present in subtelomeric regions and are subject to reversible repression. Examining the subtelomeric galactose-inducible HXK1 gene on telomere VIR, Taddei and colleagues (73) showed by confocal microscopy that this region became confined to the nuclear periphery upon galactose induction. Independent analysis showed that a similar phenomenon occurred for the GAL1 gene and that optimal gene induction was dependent on Ada2 (13).
We asked whether Ada2 telomeric localization changed in response to alternative carbon sources by assessing Ada2 and Gcn5 occupancy at the telomere by ChIP. In glucose-grown cells, Ada2 occupancy was enriched eightfold over that for untagged samples at 200 bp from telomere VIR and decreased to approximately fivefold enrichment at 5 kb from the telomere (Fig. (Fig.6,6, top). Of note, Ada2 occupancy was detected at 15 kb and 20 kb from the telomere, near the SAGA-regulated gene HXK1, even though HXK1 expression is repressed under glucose conditions. Ada2p occupancy at the rDNA array was assayed at the 25S (proximal to the ADE2-CAN1 reporter gene) and 5S genes and was approximately fivefold increased over that for untagged samples. As a control, the rDNA-associated protein Net1 (69) exhibited higher occupancy than Ada2 at the rDNA 25S and 5S genes, at 14-fold and 41-fold enrichment, respectively. Net1 occupancy was significantly lower than Ada2 occupancy at the telomere, as expected. Gcn5 exhibited cooccupancy with Ada2 at both telomere and rDNA loci.
Upon galactose induction, Ada2 occupancy was enriched 16-fold and 13.5-fold over that for untagged samples at 200 bp and 1 kb from the telomere, respectively, with a maximum Gal/Glc occupancy ratio of 2.5 at 5 kb (Fig. (Fig.6,6, bottom). Ada2 occupancy further inward, at HXK1, was elevated 1.8-fold over that under glucose-grown conditions. Ada2 association with the rDNA array likewise increased, with Gal/Glc occupancy ratios of 3.7 and 3.2 at the 25S and 5S genes, respectively. Net1p occupancy at the 5S gene and at the telomere locations surveyed did not increase upon growth in galactose. The Gal/Glc ratio of Ada2 occupancy at the positive-control GAL1 promoter was the highest observed, at 4.7. Although the levels of Ada2 and Gcn5 occupancy at telomeres differed at different telomeric positions, they were comparable at the GAL1 promoter, which is subject to Gcn5-dependent activation upon growth in galactose (10). Thus, the profile of Ada2 telomeric occupancy is sensitive to nutrient changes, with the maximal binding shifting inward, to 5 kb from the telomere.
To assess Ada2 occupancy at an independent telomere, ChIP samples were analyzed with primers directed toward telomere IXR at 2 kb, 4 kb, and 9 kb from the terminus (see Fig. S6 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm). Ada2 occupancy was maximal at 2 kb from the terminus, at 6.5-fold over the level for untagged control samples. Its occupancy decreased in a manner similar to that for telomere VIR for points surveyed at 4 kb and 9 kb. Upon a shift to galactose, cells exhibited increased Ada2 occupancy at each location, with the maximal Gal/Glc ratio at 4 kb from the telomere end, similar to that observed with telomere VIR. Thus, inward shifting of maximal Ada2 occupancy upon growth in galactose was observed in two telomeres with different architectures.
Given that Ada2 is important for telomeric silencing and that growth in galactose changed the Ada2 occupancy profile to maximal occupancy at the 5-kb region of telomere VIR, we tested whether growth on galactose affects telomeric silencing in wild-type cells.
Indeed, wild-type cells grown on galactose plates displayed a marked telomeric silencing defect relative to wild-type cells grown on glucose plates (Fig. (Fig.7A).7A). This silencing defect was not due to decreased levels of Ada2 in cells grown in galactose (Fig. (Fig.7B),7B), nor was it dependent on GCN5 (see Fig. S7b posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm). The galactose-induced telomeric silencing defect was also assessed by an independent approach. The mRNA levels of the telomere VIR YFR057W locus, spanning 1 to 2 kb from the telomere, were examined by RT-PCR. YFR057W RNA levels increased 3.5-fold upon the shift to galactose growth, a level comparable to that observed in a sir2Δ control strain (Fig. (Fig.7C).7C). Thus, loss of telomeric silencing upon a nutrient shift correlates with an Ada2 occupancy shift at the telomere.
Because growth in an alternative carbon source correlated with a telomeric silencing defect, we tested whether telomeric Sir occupancy and H3 K14 acetylation levels changed under these conditions. Telomeric Sir3 occupancy did not change significantly upon a shift to galactose (Fig. (Fig.7D).7D). Likewise, H3 K14 acetylation was also unchanged in wild-type cells grown in galactose (see Fig. S7a at the URL mentioned above). The finding of constant H3 K14 acetylation in wild-type cells yet increased H3 K14 acetylation in sir3Δ cells is consistent with the data presented in Fig. Fig.5B5B (bottom), in which the four- to fivefold increase in H3 K14 acetylation at 200 bp and 1 kb from the telomere occurred only upon loss of Sir3.
Based on the observation that growth in an alternative carbon source altered the Ada2 telomeric occupancy profile but did not reduce overall Ada2 abundance, we tested whether an increased gene dosage of ADA2 would rescue the nutritionally induced silencing defect in wild-type cells. Indeed, significantly increased Ada2 levels improved telomeric silencing in galactose (Fig. (Fig.7E,7E, bottom right). Further, this improvement was SAGA dependent, as demonstrated by the lack of rescue in an spt7-400 strain (Fig. (Fig.7E).7E). The restoration of silencing was specific in that the sas2Δ telomeric silencing defect was not affected by increasing ADA2 expression (Fig. (Fig.7E)7E) either in glucose (left) or in galactose (right).
Thus, a telomeric silencing defect dependent on Ada2 function can be induced in two ways: by a complete loss of Ada2, which correlates with the spreading of Sir proteins inward from the telomere and a loss of boundary function, or by a change in the nutrient source, which correlates with an inward shifting of Ada2. This inward shifting of chromatin-associated and chromatin-modifying proteins may be an integral part of the switch from a repressed to an activated state of gene expression.
The observations reported here define a new function for Ada2 in promoting transcriptional silencing at telomeres and the rDNA array that is distinct from its transcriptional activation role with Gcn5. Ada2 physically associates with transcriptionally repressive heterochromatin in the context of SAGA or a related HAT complex. At the telomere, Ada2 restricts the inward spread of heterochromatic proteins in a manner that is functionally similar to, but mechanistically distinct from, that of Sas2. In physiological response to changes in nutrients, Ada2 occupancy increases at subtelomeric regions, an increase that correlates with a telomeric silencing defect yet leaves intact molecular hallmarks of silenced chromatin.
The model in Fig. Fig.88 depicts the data presented here in the context of independent studies revealing convergent functions of factors in regulating gene expression near the nuclear periphery (for a review, see references 1 and 2). Here, three important aspects of regulation are linked: positioning of heterochromatic regions near the nuclear periphery, heterochromatin-euchromatin boundary formation, and the conferring of optimal gene expression through nuclear pore association.
Our studies show that Ada2, likely in the context of SAGA, is positioned to promote silencing and provide boundary function and is then repositioned to mediate a switch to activated transcription of target genes. Such changes in Ada2 occupancy upon a nutrient shift from glucose to galactose are not specific for telomere VIR (Fig. (Fig.6)6) but also occur at another telomere, IXR (see Fig. S6 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm), at which no Gal-inducible genes are known. Thus, the Ada2 occupancy increase at subtelomeric locations is independent of the telomere identity or the presence of a Gal-regulated gene. It may, considering SAGA's role in sensing environmental conditions, be a more general response to nutrient utilization and stress conditions.
There are several possibilities for the mechanism by which Ada2 promotes telomeric silencing. Two are considered here. The first is that Ada2 could directly or indirectly block the HAT activity of Gcn5. This idea is based on the observation that H3 K14 is normally hypoacetylated near the telomere (9, 56) but becomes hyperacetylated in a silencing-defective sir3Δ strain (Fig. (Fig.5B).5B). Further, Gcn5 directly tethered to the telomere can disrupt silencing (16, 33, 34). However, two observations make this activity-blocking explanation unlikely. The ada2Δ and sir3Δ strains are comparably defective in telomeric silencing, but an ada2Δ strain does not exhibit increased H3 K14 acetylation at the chromosomal terminus (Fig. (Fig.5B).5B). Also, deletion of GCN5 in an ada2Δ background does not suppress the ada2Δ telomeric silencing defect, as would be predicted if Gcn5 HAT activity were responsible for the defect (data not shown).
A second possibility is that Ada2, perhaps in conjunction with another SAGA subunits, acts as a physical block to the spread of heterochromatin proteins inward from the telomere. This idea is supported by the observation that mutation of some, but not all, SAGA subunits disrupts telomeric silencing (Fig. (Fig.1C).1C). A physical blocking model has been proposed for the boundary activity of the mammalian DNA binding transcription factor Sp1 (33) and nuclear pore-associated exportins (32) at HML, and for the RNA polymerase III complex at the downstream barrier tRNA gene at HMR (20).
The physical location of Ada2 at the telomere and its co-IP with Sir2 support a physical blocking model. This interaction may be mediated through intermediary binding partners or DNA at telomeres and/or the rDNA, and it indicates that a fraction of SAGA is associated with the nuclear periphery. Telomeric heterochromatin associates with the nuclear periphery through a Sir-dependent interaction between Sir4 and Esc1, and in a Sir-independent manner through Ku80 (reviewed in reference 25). Ada2 and SAGA may be spatially proximal to nuclear envelopes through interactions with nuclear pore components, although telomeric foci are not coincident with nuclear pores (reviewed in reference 25). Previous studies have demonstrated that Mlp1 of the inner nuclear basket associates with SAGA components Gcn5, Ada2, and Spt7 at the GAL1 promoter, and this interaction depends on the integrity of the SAGA complex (39). Likewise, the nuclear pore protein Nup2 interacts with the SAGA-dependent genes GAL1 and HXK1 upon Gal induction (61), and Nup2 can provide boundary activity upon targeting to a silenced locus (32).
The studies presented here highlight distinctions in boundary formation and heterochromatin-mediated silencing. Although the ada2Δ and sas2Δ telomeric silencing defects have shared characteristics of Sir spreading and decreased subtelomeric H4 K16 acetylation (Fig. (Fig.44 and and5),5), they differ mechanistically. As shown in Fig. S4 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm, the ada2Δ but not the sas2Δ telomere silencing defect is rescued by an increased dosage of SIR3; the sas2Δ silencing defect is modestly and selectively rescued by increased SIR4 dosage; and Ada2 and Sas2 do not reciprocally rescue each other's telomeric silencing defects. At HMR, deletion of ADA2 was not sufficient to promote Sir3 spreading: instead, disruption of both a chromatin-modifying enzyme (Gcn5) and a DNA binding factor at the tRNA gene boundary element flanking HMR (45) was necessary in order to disrupt boundaries. In contrast, at telomeres, the function of Ada2 as a physical barrier to heterochromatin spread or the chromatin-modifying activity of Sas2 is sufficient for boundary function.
The mechanism by which Ada2 localizes to telomeres remains to be elucidated. A functional silencing complex is not required (Fig. S3 posted at http://www-biology.ucsd.edu/labs/pillus/pubs.htm). Chromatin-targeting domains in SAGA subunits or Ada2 interaction with resident telomeric and subtelomeric proteins may be important for Ada2 recruitment. Resident telomeric and subtelomeric factors, including telomere-associated proteins such as Ku80, nuclear basket or pore proteins such as Mlp1 or Nup2, and transcriptional activators/repressors such as Rap1, Tup1, and Gal4, may all contribute to recruitment.
Rap1 and Tup1 are of particular interest because they also respond to nutritional cues and mediate transcriptional switches. Rap1 is an abundant DNA-binding protein that plays a central role in metabolism (63) and underpins telomere structure and Sir complex formation through the binding of its target sequences (reviewed in references 2 and 59). Tup1 represses the transcription of some subtelomeric genes (52, 56) as part of the Cyc8-Tup1 corepressor complex and, upon induction by stress or nutrient changes, can mediate gene activation through SAGA recruitment (47, 50). Tup1 and Rap1 can act coordinately in nutrient-induced transcriptional switches whereby Tup1-dependent restriction of Rap1 occupancy at low-glucose target genes is alleviated upon glucose depletion (12). Future studies to identify factors that recruit and participate with Ada2 in its telomere-silencing function will be instrumental in developing a detailed molecular mechanism of Ada2 recruitment and chromatin-dependent switches in transcriptional regulation.
Posttranslational histone modifications by or to SAGA subunits themselves may also contribute to Ada2 telomere recruitment and switches between repression and activation. Perinuclear confinement of GAL1 upon activation requires Ada2 and the nuclear export protein Sus1 (13). Sus1 physically associates with both the SAGA complex and the nuclear export complex (57). Recent evidence indicates that the central domain of Ada2 binds phosphatidylserine (29), suggesting an intriguing role for this protein in gene expression through phosphoinositide signaling.
The significance of Ada2 and other SAGA subunits in both repressed and activated transcription is likely to grow. Indeed, it may underscore developmental defects observed in D. melanogaster upon deletion of Ada2α and Ada2β (46), where the timing of gene expression is of critical importance (reviewed in reference 65). Further, telomeric silencing is one mechanism for regulating virulence genes in human pathogens (19). The paradox that perinuclear localization is necessary for repression as well as for optimal activation may be resolved simply by the increasing evidence of SAGA subunits and other telomere-associated proteins that act as transcriptional activators or repressors depending on their chromosomal binding contexts (reviewed in reference 48). Thus, further analysis of the role of Ada2 in the transcriptional regulation of telomere-proximal genes may provide mechanistic insight into the regulation of subtelomeric genes that contribute to development and disease.
We thank the members of our lab for continuing discussions; T. Laurenson, R. Darst, T. Johnson, and A. Hsu for insightful comments on the manuscript; and S. Berger, M. P. Cosma, M. M. Smith, F. Winston, F. Robert, C. Brandl, and R. Hampton for reagents.
The laboratory has been supported by the National Institutes of Health.
Published ahead of print on 8 September 2009.