• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of molcellbPermissionsJournals.ASM.orgJournalMCB ArticleJournal InfoAuthorsReviewers
Mol Cell Biol. Jan 2009; 29(1): 266–280.
Published online Oct 20, 2008. doi:  10.1128/MCB.00315-08
PMCID: PMC2612497

The STAGA Subunit ADA2b Is an Important Regulator of Human GCN5 Catalysis[down-pointing small open triangle]

Abstract

Human STAGA is a multisubunit transcriptional coactivator containing the histone acetyltransferase GCN5L. Previous studies of the related yeast SAGA complex have shown that the yeast Gcn5, Ada2, and Ada3 components form a heterotrimer that is important for the enzymatic function of SAGA. Here, we report that ADA2a and ADA2b, two human homologues of yeast Ada2, each have the ability to form a heterotrimer with ADA3 and GCN5L but that only the ADA2b homologue is found in STAGA. By comparing the patterns of acetylation of several substrates, we found context-dependent requirements for ADA2b and ADA3 for the efficient acetylation of histone tails by GCN5. With human proteins, unlike yeast proteins, the acetylation of free core histones by GCN5 is unaffected by ADA2b or ADA3. In contrast, the acetylation of mononucleosomal substrates by GCN5 is enhanced by ADA2b, with no significant additional effect of ADA3, and the efficient acetylation of nucleosomal arrays (chromatin) by GCN5 requires both ADA2b and ADA3. Thus, ADA2b and ADA3 appear to act at two different levels of histone organization within chromatin to facilitate GCN5 function. Interestingly, although ADA2a forms a complex(es) with GCN5 and ADA3 both in vitro and in vivo, ADA2a-containing complexes are unable to acetylate nucleosomal H3. We have also shown the preferential recruitment of ADA2b, relative to ADA2a, to p53-dependent genes. This finding indicates that the previously demonstrated presence and function of GCN5 on these promoters reflect the action of STAGA and that the ADA2a and ADA2b paralogues have nonredundant functional roles.

Eukaryotic DNA is packaged within a chromatin structure that has several levels of organization: the fundamental nucleosome subunit comprising 147 bp of DNA wrapped around the core histones H2A, H2B, H3, and H4 and higher-order structures resulting from internucleosomal interactions and the incorporation of linker histones (53). This arrangement results in a general repression of transcription that can be relieved by the actions of ATP-dependent chromatin-remodeling enzymes and coactivators that effect covalent modifications of the accessible histone tails (37). Covalent modifications of histones include acetylation, methylation, ubiquitination, and phosphorylation, and there is ample evidence that the addition or removal of the corresponding groups influences the interaction of chromatin with nonhistone binding partners, including transcription cofactors (33). Apart from the mechanisms involved in the recruitment of histone-modifying complexes, an understanding of the catalytic functions and substrate specificities of these complexes is fundamental for an appreciation of their cellular roles in transcription control.

Dynamic reversible acetylation of histones is generally associated with gene activation and mediated by several histone acetyltransferases (HATs) with different and sometimes overlapping specificities (36). Mammalian HATs show greater diversity than yeast HATs, potentially serving gene- or cell type-specific functions. Two of these enzymes, GCN5 (of which two alternatively spliced forms, GCN5S and GCN5L, have been described previously [58]) and PCAF, are paralogues that are related to yeast Gcn5 (yGcn5) (12, 60). Like yGcn5, which is found in the SAGA complex (23), mammalian GCN5 and PCAF are found in large complexes comprising at least a dozen subunits. These include the GCN5L-containing STAGA complex (39), the TATA-binding protein-free TAF-containing complex (TFTC) incorporating GCN5L (56), and the PCAF complex (42). The mammalian (human) complexes contain homologues of virtually all yeast SAGA subunits, as well as associated factors involved in DNA repair and RNA processing (8, 40, 63). PCAF null mice develop normally and lack a distinct phenotype. In contrast, GCN5 null embryos die during embryogenesis (57, 59). The mutation of the HAT domain of GCN5 is also lethal in mice, but at a later embryonic stage than the complete knockout of the GCN5 gene (10).

With respect to catalytic mechanisms, previous studies of Saccharomyces cerevisiae have shown that Ada proteins yGcn5, yeast Ada2 (yAda2), and yeast Ada3 (yAda3) form a heterotrimer (29). Based on the observation that yAda2 and yAda3 strongly enhance the ability of yGcn5 to acetylate nucleosomes (3), this subcomplex was suggested to form the catalytic cores of the related SAGA, ADA, and SALSA, or SLIK, complexes. In relation to the possibility of a similar core subunit in higher eukaryotes, two human (and Drosophila) homologues of yAda2 have been described previously (4, 34). Human ADA2a was found to interact with human GCN5, and ectopic ADA2a or ectopic GCN5 was found to enhance (albeit not synergistically) GAL4-VP16-dependent transcription in luciferase assays (12). ADA2a was subsequently identified in the PCAF complex (42) but not in TFTC (4, 9). Human ADA2b was identified as a PCAF-interacting protein in a yeast two-hybrid screen and was also found to coprecipitate with GCN5 and BRG1 from HeLa cells (4). However, ADA2b was reported not to coprecipitate with overexpressed cellular PCAF and is absent in the GCN5L-containing TFTC and the BRG1-containing human SWI/SNF complex (4). The results of functional assays showed previously that ectopic human GCN5 and ADA2b act synergistically to enhance the PAX5-dependent activation of an ectopic expression vector (4).

To analyze how human STAGA subunits modulate the catalytic activity of GCN5L, we compared the activity of isolated recombinant GCN5L both to that of the STAGA complex and to those of trimeric complexes composed of recombinant GCN5L, ADA3, and either ADA2a or ADA2b. We have shown that ADA2b, unlike ADA2a, enhances the acetylation of nucleosomes by GCN5L and, using ADA2 paralogue-specific antibodies, that STAGA contains ADA2b but not ADA2a. We characterized the interactions governing GCN5-ADA2b-ADA3 heterotrimer formation and, by comparing the patterns of acetylation of core histones, mononucleosomes, and nucleosomal arrays (chromatin), also demonstrated a context-dependent requirement for ADA2b and ADA3 for the efficient acetylation of histone tails by GCN5. In agreement with our previous findings that STAGA interacts with, and is recruited by, p53 upon DNA damage (22), we have shown that ADA2b but not ADA2a is recruited to p53-dependent promoters after UV damage.

MATERIALS AND METHODS

Cell culture and transfection.

U2OS cells were kept in Dulbecco modified Eagle medium supplemented with 10% fetal calf serum. HeLa cells in spinner cultures were kept in phosphate-buffered Dulbecco modified Eagle medium with 10% bovine serum for large-scale nuclear extract preparation. Stable cell lines were created as described previously (40). Transfections were conducted using FuGENE6 (Roche) for U2OS cells and Lipofectamine (Invitrogen) for HeLa cells according to the instructions of the reagent manufacturers. The HeLa cell line that stably expresses FLAG-hemagglutinin (HA)-tagged PCAF was kindly provided by Yoshihiro Nakatani.

Cloning of STAGA subunits and antibodies.

The cDNAs for human ADA2a and ADA2b were obtained from expressed sequence tags. Human ADA3 cDNA was cloned from a fetal spleen cDNA library. Antibodies against ADA2b and ADA2a were raised in rabbits injected with bacterially purified His-tagged ADA2b and a glutathione S-transferase (GST)-fused mutant form of ADA2a comprising amino acids 296 to 443 [ADA2a(296-443)]. Antibodies against TRRAP, GCN5, actin, and the large subunit of RNA polymerases II (N20) were from Santa Cruz. The antibodies against ADA3, TAF5L/PAF65β, and TAF9 were kindly provided by Yoshihiro Nakatani and Laszlo Tora.

Nuclear extract preparation and complex purification from stable cell lines.

Nuclear extract from HeLa cells was prepared as described previously (17). Nuclear extract from U2OS cells was prepared on a small scale (46). The STAGA and PCAF complexes were purified from the nuclear extracts of HeLa cell lines that stably express FLAG-HA-tagged SPT3 (40) or PCAF (42) by immunopurification on M2 agarose (Sigma), three washes with BC buffer (10% glycerol, 20 mM Tris-HCl, pH 7.9, 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride [PMSF], 1 mM dithiothreitol [DTT], 0.05% NP-40) containing 300 mM KCl, and one wash with BC buffer containing 100 mM KCl. The complexes were eluted with FLAG peptide in BC buffer containing 100 mM KCl. ADA3-, ADA2b-, and STAF65γ-interacting proteins from HeLa cells and ADA2a- and ADA2b-containing complexes from U2OS cells were purified analogously.

Purification of histones and nucleosomes and assembly of nucleosomal arrays.

HeLa cell nuclear pellets were obtained as residues from nuclear extract preparation (17) and resuspended in buffer I (0.1 M potassium phosphate, pH 6.7, 0.63 M NaCl, 10% glycerol, 0.5 mM EDTA, 1 mM DTT, 0.5 mM PMSF) by homogenization with a size B Dounce pestle. After centrifugation, the DNA concentration in the supernatant was adjusted to 1 mg/ml with buffer I, and the mixture was rotated with hydroxyapatite (Bio-Rad) for 2 h at 4°C. The resin was collected and washed four times with buffer I. After resuspension in buffer I, the slurry was poured into a Bio-Rad column and washed once more with buffer I. The column was slowly eluted with buffer B (0.1 M potassium phosphate, pH 6.7, 2 M NaCl, 10% glycerol, 0.5 mM EDTA, 1 mM DTT, 0.5 mM PMSF). Fractions were stored in buffer B at 4°C and checked by Coomassie blue staining. The obtained core histone preparation was both concentrated and free of HAT activity. Nucleosomes were purified from HeLa cell nuclear pellets according to a protocol described previously (15). Fractions 12 and 14 served as mononucleosomes with or without a linker histone (see Fig. S1 in the supplemental material). Nucleosomal arrays were assembled from unmodified, bacterially purified recombinant histones as described previously (1).

HAT assays.

HAT assays were carried out at 30°C as described previously (39). Two ways of analyzing tritium-labeled acetyl group transfer from [3H]acetyl coenzyme A ([3H]AcCoA; Amersham) to proteins were used: the separation of proteins by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by Amplify reagent (Amersham Pharmacia)-enhanced fluorography analysis of dried gels, and filter binding assays (39) that took advantage of the phosphocellulose binding property of histones. The radioactivity on washed P81 filters (Whatman) was measured by a scintillation counter. Where indicated, results were obtained from three independent experiments. Filter binding assays were used for kinetic analyses. Initial rates (acetyl transfers per GCN5 molecule per second) were determined by varying the concentrations of histones and mononucleosomes. The histones used in this particular experiment were in the form of a histone mixture (product no. H9250; Sigma). Unlike the histones purified by the protocol described above, this histone preparation had residual HAT activity that had to be corrected for in the assays. GCN5 concentrations were 5 or 10 nM, and the AcCoA concentration was 5.6 μM. Histone saturation curves were fitted to the Michaelis-Menten equation by using the program Prism 5 (GraphPad Software).

Purification of proteins.

FLAG-tagged GCN5L and GCN5S were purified from Sf9 insect cells. M2 agarose-bound protein was washed three times with BC buffer containing 400 mM KCl and once with BC buffer containing 100 mM KCl before elution with FLAG peptide. Because bacterially expressed ADA2b was found nearly exclusively in the insoluble fraction, the two human ADA2 paralogues were purified from insect cells as dimers with GCN5L. For the copurification of GCN5L-ADA2a or GCN5L-ADA2b heterodimers, Sf9 cells were coinfected with viruses for the untagged protein and the FLAG-tagged subunit, with the baculovirus for the untagged protein at an excess titer compared to that of the virus for the FLAG-tagged subunit. Proteins in derived extracts were loaded onto M2 agarose in a mixture of 20 mM Tris-HCl, pH 7.9, 15% glycerol, 0.01% NP-40, 250 mM NaCl, 2 mM MgCl2, 0.2 mM EDTA, and 0.5 mM PMSF. The M2 agarose beads were washed four times with a mixture of 20 mM Tris-HCl, pH 7.9, 15% glycerol, 0.01% NP-40, 150 mM NaCl, 2 mM MgCl2, 0.2 mM EDTA, and 0.5 mM PMSF and once with BC buffer containing 100 mM KCl before elution with BC buffer containing 100 mM KCl and 0.5 μg/μl FLAG peptide. In order to compare the enzymatic activity of FLAG-tagged GCN5L to those of the heterodimers, the non-dimer-associated protein used in the assays was purified in parallel in the same way as the heterodimers. Because washing conditions were less stringent than those used for the isolated individual GCN5 polypeptides, certain contaminants were observed. Full-length and deletion mutant ADA3 proteins were purified from bacteria as His-tagged or FLAG-tagged proteins.

In vitro interaction assays.

GST fusion proteins were expressed in Escherichia coli strain XA90 and purified as described previously (25). Proteins were labeled with [35S]methionine by using TNT reticulate lysates according to the standard protocol provided by Promega. In vitro-translated proteins were incubated with resin-bound proteins (10 μg) with rotation at 4°C for 3 to 4 h in BC buffer containing 150 mM KCl and 0.1 μg/μl bovine serum albumin. After washes with BC buffer containing 200 mM KCl, the resin was boiled in SDS loading buffer before being subjected to SDS-PAGE. Gels were incubated in Amplify (Amersham Pharmacia) before being dried and subjected to autoradiography to enhance the signal.

ChIP assays.

Chromatin immunoprecipitation (ChIP) assays were performed as described previously (1). UV-treated U2OS cells were irradiated with 50 J/m2 and harvested after 6 h. Quantitative PCR was used to evaluate the results of at least three independent ChIP experiments. Antibodies against HA (clone 12CA5) were kindly provided by the lab of David Allis.

RNA interference.

U2OS cells were transfected with small interfering RNAs directed against human ADA2a or nontargeting control small interfering RNAs (all from Dharmacon). Forty-eight hours after transfection, cells were irradiated with UV and harvested after 8 h for RNA isolation with TRIzol (Invitrogen). RNA was reverse transcribed with SuperScript II reverse transcriptase according to the protocol of the manufacturer (Invitrogen).

RESULTS

Subunits of the STAGA complex modify the acetyltransferase activity of its catalytic subunit GCN5L.

It has been reported previously that the human PCAF and yGcn5 enzymes are less efficient in acetylating nucleosomes than the corresponding multisubunit complexes (23, 42). Also, regarding the long amino-terminal regions of PCAF and GCN5L that are absent in both yGcn5 and the short mammalian GCN5 splice variant (human and murine GCN5S), the long forms of GCN5 are more active than the short forms in acetylating nucleosomes (58, 60). These observations warranted a further investigation of the modulation of human GCN5 catalytic activity by amino-terminal sequences and associated subunits.

For these analyses, STAGA was purified from a cell line that stably expresses FLAG-HA-tagged SPT3 (Fig. (Fig.1A).1A). GCN5L and GCN5S were purified from Sf9 cells as FLAG-tagged proteins (Fig. 1 B). Comparative Coomassie blue staining and immunoblotting were used to estimate the molar amounts of complex-associated and recombinant GCN5. With free histone substrates, GCN5L, GCN5S, and STAGA acetylated mainly H3 and, to a lesser degree, H4 (Fig. (Fig.1C).1C). Interestingly, GCN5L acetylated core histones about fivefold more efficiently than GCN5S, indicating that the GCN5L amino terminus, which has no counterpart in yGcn5, increases free core histone acetylation as well as nucleosome acetylation (58, 60). The level of enzymatic activity of the STAGA complex on free histones was only slightly higher than that of recombinant GCN5L (Fig. (Fig.1C).1C). This finding is somewhat surprising in light of the fact that yAda2, which forms a heterotrimer with yAda3 and yGcn5 (29), was reported previously to increase the ability of yGcn5 to acetylate core histones (3). This result may indicate some special property of the amino-terminal sequences found in human GCN5L but not in yGcn5.

FIG. 1.
STAGA subunits increase the ability of GCN5 to acetylate nucleosomes. (A) Analysis of purified STAGA. The complex was purified from a HeLa cell line that stably expresses FLAG-HA-tagged SPT3 (fha:SPT3) and analyzed by SDS-PAGE with silver staining. M, ...

Next, we compared the abilities of GCN5L and STAGA to acetylate mononucleosome substrates by using a filter binding assay to quantitate the activities. When comparable molar amounts were assayed, STAGA was found to be at least 20 times more efficient than GCN5L in acetylating mononucleosomes (Fig. (Fig.1D).1D). This outcome suggests that the nucleosomal environment of the histones elicits a requirement for additional subunits for efficient histone acetylation by GCN5L.

Human ADA2a and ADA2b interact with GCN5 and ADA3 in vivo and in vitro.

We next tested whether STAGA contains a catalytic core analogous to the yGcn5-yAda2-yAda3 heterotrimer (3). While the main difference between yGcn5 and human GCN5L is the presence of a long amino-terminal domain in the latter, the ADA3 subunit (49 kDa) of STAGA (40) is much smaller than yAda3 (79 kDa) and lacks homology to several noncontiguous regions. The two human ADA2 homologues are similar in size and have 26% identity and 50% homology in amino acid sequences. Like yAda2, they both have zinc finger, SANT, and SWIRM domains (Fig. (Fig.1E),1E), and they are similar in size to yAda2. The relative roles of the two ADA2 homologues in mammalian complexes are yet unknown, and as mentioned in the introduction, current reports concerning their incorporation into complexes are somewhat contradictory.

To see which ADA2 homologue is able to form a complex with GCN5L, we tested intracellular interactions with endogenous GCN5L by transient expression of FLAG-HA-tagged ADA2a or ADA2b in H1299 cells. Derived whole-cell lysates were incubated with M2 (anti-FLAG) agarose, and immunoprecipitates were analyzed by immunoblotting. Endogenous GCN5L was coimmunoprecipitated with both ADA2a and ADA2b, albeit more efficiently with the latter (Fig. (Fig.2A,2A, compare lanes 2 and 3). To test whether the interaction of the ADA2 proteins with GCN5L is direct, GST and GST-fused full-length ADA2 proteins were independently bound to glutathione-Sepharose and incubated with in vitro-translated GCN5L. GCN5L bound to both full-length ADA2 paralogues but slightly more strongly to ADA2b (Fig. (Fig.2B,2B, upper panel). A similar analysis of in vitro-translated ADA3 showed comparable levels of binding to the two ADA2 proteins (Fig. (Fig.2B,2B, lower panel).

FIG. 2.
ADA2a and ADA2b both interact with GCN5 and ADA3, but only ADA2b is a subunit of STAGA. (A) Coprecipitation of endogenous GCN5L with FLAG-HA-tagged ADA2a (fha:ADA2a) or ADA2b overexpressed in H1299 cells. Inputs (IN) and anti-FLAG (M2 agarose) immunoprecipitates ...

ADA2b, but not ADA2a, is a component of the STAGA complex.

ADA2a-GCN5 and ADA2b-GCN5 heterodimers were expressed in and purified from Sf9 cells (Fig. (Fig.2C)2C) and employed to show that antibodies directed against GST-ADA2a(296-443) and His6-tagged ADA2b were specific, respectively, for ADA2a and ADA2b in the heterodimers (Fig. (Fig.2E,2E, compare lanes 2 and 3). The high-level specificities of the antibodies allowed a comparison of the STAGA and PCAF complexes that were purified from HeLa cell lines that stably express FLAG-HA-tagged SPT3 and PCAF, respectively. ADA2b was present in both preparations, which were normalized according to the common TRRAP subunit, whereas ADA2a was observed only in the PCAF preparation (Fig. (Fig.2E,2E, compare lanes 4 and 5). Although it is clear that the STAGA complex contains only the ADA2b paralogue, one possible explanation for the presence of both ADA2 paralogues in the PCAF preparation is that this preparation contained more than one PCAF complex—one identical to STAGA except for the presence of PCAF instead of GCN5L and another(s) containing PCAF and ADA2a.

To test this hypothesis, we analyzed M2 immunoprecipitates from HeLa cell lines that stably express the FLAG-HA-tagged STAGA subunit ADA3, ADA2b, or STAF65γ. Immunoblots for both ADA2a and several STAGA subunits revealed that all previously identified and tested STAGA subunits coimmunoprecipitated with ADA3 (Fig. (Fig.3A).3A). Interestingly, however, and in agreement with the direct in vitro interaction of both human ADA2 paralogues with human ADA3 (Fig. (Fig.2B),2B), ADA2b and ADA2a were both coimmunoprecipitated with ADA3 (Fig. (Fig.3A).3A). Thus, apart from being a subunit of STAGA, ADA3 is also found in an ADA2a-containing complex(es) that is potentially similar to the Drosophila ATAC complex (26). In order to test whether ADA2b and ADA2a are found in exclusive complexes, we compared immunoprecipitates from cell lines expressing FLAG-tagged ADA3, ADA2b, SPT3, or STAF65γ. All tested STAGA subunits, including SPT3, TRRAP, GCN5L, PAF65β/TAF5L, and ADA2b, coimmunoprecipitated with all four tagged proteins (Fig. (Fig.3B).3B). Because STAGA constitutes only a fraction of ADA3-containing complexes, the STAGA subunit signals in the FLAG-ADA3 immunoprecipitate were weaker than those in the other precipitates. ADA2a was coimmunoprecipitated only with ADA3 and, importantly, was absent in the ADA2b, SPT3, and STAF65γ immunoprecipitates. Hence, there exist at least one human ADA3-containing complex and, by inference from the PCAF immunoprecipitate, at least one GCN5-containing complex besides STAGA. These complexes contain ADA2a, although it remains to be seen whether all ADA2a-containing complexes contain both ADA3 and GCN5.

FIG. 3.
ADA2a and ADA2b are exclusive components of distinct complexes. (A) Presence of ADA3 in both STAGA and an ADA2a-containing complex(es). An anti-FLAG (M2 agarose) immunoprecipitate from a HeLa cell line that stably expresses FLAG-HA-tagged ADA3 (fha:ADA3) ...

For further confirmation of these results, we also generated U2OS cells that stably expressed FLAG-HA-tagged ADA2a or ADA2b. The immunoblotting of immunoprecipitates from these cells confirmed that ADA2b and ADA2a are exclusive subunits of distinct complexes (Fig. (Fig.3C).3C). Thus, while ADA2a and ADA2b both form complexes with ADA3 and GCN5, ADA2a does not associate with several proteins (e.g., TRRAP and TAF5L/PAF65β) identified previously (40) as STAGA subunits.

To further characterize ADA3-containing complexes, we subjected the immunoprecipitate from the stable FLAG-ADA3 cell line to gel filtration on a Superose 6 column. Fractions were tested for ADA2a and various STAGA subunits by immunoblotting (Fig. (Fig.3D).3D). The majority of the ectopic FLAG-ADA3 was not associated with STAGA (present predominantly in fractions 4 to 6) but was found in lower-molecular-mass fractions (predominantly fractions 7 to 10). ADA2b was found in SPT3-, TRRAP-, and GCN5L-containing fractions, in agreement with its status as a subunit of STAGA. ADA2a was found in apparent ADA3-containing complexes with low molecular masses, in further support of its absence from the STAGA complex. A weak signal in the fraction corresponding to a size above 2 MDa suggested the existence of another high-molecular-mass complex containing ADA2a, ADA3, and GCN5L. In the absence of an available GCN5L cell line, we used a cell line that stably expressed FLAG-tagged GCN5S to purify GCN5S-associated proteins. The GCN5S immunoprecipitate was subjected to Superose 6 fractionation (Fig. (Fig.3E).3E). As expected, GCN5S was found to be associated with TRRAP, ADA3, ADA2b, and SPT3 in a STAGA-like complex in fractions 4 to 6. The composition of fraction 8 indicated the presence of an SPT3-free complex containing GCN5S, ADA3, and ADA2b. It is presently unknown whether such a complex is physiologically relevant or the product of the artificial overexpression of GCN5S, although it may reflect the stabilization of a catalytic core intermediate (see below) in STAGA formation. No ADA2a signal was detected, reflecting the previously observed selective association of GCN5 with ADA2b over ADA2a.

Despite their common ability to associate with GCN5L and ADA3 in vitro and in the cell, the ADA2 paralogues thus form distinct complexes with strikingly different compositions, suggesting diverse cellular roles.

ADA2b stimulates the acetylation of nucleosomes, but not the acetylation of core histones, by GCN5L, whereas ADA2a has no effect on either GCN5 activity.

To test the relevance of GCN5 interactions with ADA2 proteins for GCN5 catalytic activity, increasing molar amounts of free GCN5L or the GCN5L-ADA2a or GCN5L-ADA2b heterodimer were used to acetylate free core histones (Fig. (Fig.4A).4A). No significant influence of ADA2a or ADA2b on the ability of GCN5 to acetylate core histones was observed. In contrast, ADA2b, but not ADA2a, strongly enhanced the acetylation of mononucleosomes by GCN5 (Fig. (Fig.4B,4B, compare lanes 2, 3, and 6). The addition of ADA3 had no significant effect on the activity of either GCN5 alone or the GCN5L-ADA2b heterodimer in this assay (Fig. (Fig.4B,4B, compare lanes 4 and 5 with lanes 2 and 3). It also had no effect on the inert GCN5L-ADA2a heterodimer (Fig. (Fig.4B,4B, lanes 6 and 7). The STAGA subunit ADA3 therefore seems to be unimportant for the acetylation of mononucleosomes by the GCN5L-ADA2b heterodimer. The differential influence of ADA2a and ADA2b on mononucleosome acetylation by GCN5L was confirmed by a filter binding assay (Fig. (Fig.4C,4C, left side). We also tested mononucleosomes as GCN5 substrates in the presence of the linker histone H1 (Fig. (Fig.4C,4C, insert), which had no significant effect on the ADA2b-independent activity of GCN5 on free histones in this assay but inhibited the acetylation of mononucleosomes by the GCN5-ADA2b heterodimer (Fig. (Fig.4C,4C, right side). A similar repressive effect mediated by H1 on the acetylation of oligonucleosomes by either recombinant PCAF or the PCAF complex was reported previously (28).

FIG. 4.
ADA2b, but not ADA2a, increases the acetylation of nucleosomes by GCN5L. (A) Lack of influence of ADA2a and ADA2b on the acetylation of free core histones by GCN5L. HAT assays with free core histones from HeLa cells were conducted with increasing amounts ...

In order to see whether cellular complexes formed by ADA2a and ADA2b showed acetylation capacities similar to those of the isolated ADA2-GCN5L heterodimers, we compared M2 agarose immunoprecipitates from U2OS cells that stably expressed FLAG-HA-tagged ADA2a or ADA2b. Because ADA2b precipitated more GCN5L than did ADA2a (Fig. (Fig.3C),3C), as in the case of transient expression in H1299 cells (Fig. (Fig.2A),2A), the two immunoprecipitates were normalized according to their free core histone acetylation activities (Fig. (Fig.4D,4D, lower panel, lanes 3 and 4), which were earlier (Fig. (Fig.4A)4A) correlated with their GCN5 levels. Normal U2OS cell extracts (Fig. (Fig.4D,4D, lane 2) or buffer (Fig. (Fig.4D,4D, lane 1) served as a negative control. In agreement with the results for recombinant GCN5-ADA2 heterodimers, mononucleosomes were strongly acetylated by the FLAG-ADA2b immunoprecipitate but only very weakly by the FLAG-ADA2a immunoprecipitate (Fig. (Fig.4D,4D, upper panel, lanes 3 and 4).

In light of the evidence that yAda2 enhances the acetylation of core histones by yGcn5 (3), we compared the enzymatic activities of human GCN5 and GCN5-ADA2b in a steady-state kinetic analysis to confirm our results. A mixture of calf thymus histones was used in this assay. Amounts of GCN5 in the free form and in the form of a GCN5-ADA2b heterodimer were estimated by Coomassie blue staining, and equimolar amounts in the assay mixtures were confirmed by quantitative immunoblotting (Fig. (Fig.4E,4E, right panel; see also Fig. S2A in the supplemental material). Initial velocities (in acetyl transfers per molecule of GCN5 per second) at various histone concentrations were determined by filter binding assays, and the data were analyzed by being fit to a Michaelis-Menten equation (Fig. (Fig.4E,4E, left plot). Our assumption that H3 is the only substrate seems to be justified by the substrate specificity of GCN5 (Fig. (Fig.1C).1C). Similar Km and Vmax values for human GCN5 (1.1 ± 0.42 μM and 0.037 ± 0.0035 s−1) and the GCN5-ADA2b heterodimer (1.0 ± 0.47 μM and 0.028 ± 0.0031 s−1) were observed. Free GCN5L and the GCN5-ADA2b heterodimer therefore have similar affinities for free histone H3 and similar catalytic turnover rates. To efficiently acetylate free histones, GCN5L thus does not require an ADA2 SANT domain for recognizing H3 tails or for undergoing a conformational change, as was proposed previously for its shorter yeast homologue (7). However, human GCN5 and GCN5-ADA2b displayed very different catalytic rates for the acetylation of mononucleosomes (Fig. (Fig.4E,4E, right plot), confirming the importance of ADA2b in that reaction.

ADA3 and GCN5L interact with distinct ADA2b domains.

In order to better understand the functional interactions of GCN5L, ADA2b, and ADA3, we further analyzed their physical interactions. To determine the regions of ADA3 necessary for ADA3 binding, in vitro-translated deletion mutant forms of ADA3 were incubated with glutathione-Sepharose-bound GST-ADA2b. An analysis of bound proteins showed that the carboxy-terminal 132 amino acids (301 to 432) of ADA3 were sufficient for binding to ADA2b (Fig. (Fig.5A)5A) and that the ADA3 amino terminus showed no binding. To determine the regions of ADA2b necessary for ADA3 and GCN5 binding, in vitro-translated ADA3 and GCN5L were incubated with immobilized GST-fused deletion mutant forms of ADA2b (see Fig. S2B in the supplemental material). The results showed that ADA3 interacts independently with both a central region (residues 120 to 240) and the SWIRM domain (residues 345 to 420) of ADA2b (Fig. (Fig.5B).5B). Importantly for later experiments, the ADA2b SWIRM domain, although able to bind ADA3 on its own, was not required for the interaction of ADA2b with ADA3 since an ADA2b mutant protein lacking the SWIRM domain (and comprising residues 1 to 345) bound ADA3 as well as the full-length ADA2b did.

FIG. 5.
ADA3 and GCN5L interaction sites on ADA2b. (A) Interaction of the carboxy terminus of ADA3 with ADA2b. Beads with GST alone or GST-fused ADA2b were incubated with various in vitro-translated amino-terminal deletion mutant ADA3 proteins, and bound proteins ...

A similar analysis of GCN5 binding to ADA2b revealed strong binding to an amino-terminal fragment (residues 1 to 120) of ADA2b but only a very weak interaction with a large carboxy-terminal fragment (residues 120 to 345) and a small amino-terminal fragment (residues 1 to 50) containing the zinc finger (Fig. (Fig.5C).5C). Thus, it seems that the SANT domain, encompassing the region from residues 51 to 120, is the main ADA2b interaction site for GCN5L. Importantly, the ADA2b interaction sites for GCN5 and ADA3 are not overlapping, and like the yeast protein (29), ADA2b thus can serve as a bridge between GCN5L and ADA3 (Fig. (Fig.5D5D).

In yeast, the deletion of amino acids 97 to 106 of the SANT domain fails to affect the binding of yAda2 to yGcn5 but abolishes the stimulatory effect of yAda2, both in the SAGA complex and in a heterodimer with yGcn5, on yGcn5 activity (7). However, the further deletion of amino acids 89 to 114 leads to the inability of yAda2 to bind yGcn5 and to the loss of yGcn5 from the SAGA complex. The deletion of the zinc finger alone was reported previously to prevent yAda2-yGcn5 heterodimer formation (11, 50) but does not lead to the loss of yGcn5 (or yAda2) from the SAGA complex (50). The yAda2 region important for interaction with yAda3 has been mapped to residues 201 to 301 of yAda2 (11). Interestingly, amino acids 200 to 228 of yAda2 show 31% identity and 59% similarity to ADA2b(201-229) and 41% identity and 76% similarity to ADA2a(203-231). Based on this high degree of conservation (the overall identity in this region among the three homologues yAda2, ADA2b, and ADA2a is 31%) and the results of our interaction studies, as well as those with yeast, this short region is likely to be a binding site for ADA3. Thus, the binding patterns for STAGA seem to be similar to the ones reported for the yeast SAGA core. We additionally observed an interaction of the ADA2b SWIRM domain with ADA3 (Fig. (Fig.5D5D).

A nucleosomal array (chromatin) substrate for GCN5 reveals a role for ADA3.

In order to use a more physiological substrate than histones and mononucleosomes for our acetyltransferase studies, we assembled nucleosomal arrays as described previously (1). We compared the patterns of acetylation of histones in the context of mononucleosomes versus nucleosomal arrays. As seen before, ADA2b stimulated the acetylation of mononucleosomes by GCN5L but ADA3 had no significant effect (Fig. (Fig.6A,6A, compare lanes 2, 3, 4, and 8; see also Fig. S3 in the supplemental material). As indicated by the upper labeled bands in Fig. Fig.6A,6A, ADA3 itself is a GCN5L substrate. (The gel was overexposed in this analysis to emphasize ADA3 acetylation, and as a result, the ADA2b-mediated enhancement of mononucleosome acetylation by GCN5 appears less pronounced than that described above.) All ADA3 deletion mutant proteins tested (Fig. (Fig.6C)6C) were able to interact with ADA2b (Fig. (Fig.5A).5A). The interacting ADA3 carboxy terminus itself was acetylated by GCN5L (Fig. (Fig.6A,6A, lane 7).

FIG. 6.
Effect of ADA3 deletions on ADA3 and nucleosomal histone acetylation by the GCN5L-ADA2b heterodimer. (A) Results of HAT assays with mononucleosome substrates. Mononucleosomes were incubated with GCN5L or a GCN5L-ADA2b heterodimer in the absence (−) ...

In contrast to mononucleosomes, chromatin was a poor substrate both for GCN5L and for the GCN5L-ADA2b heterodimer (Fig. (Fig.6B,6B, lanes 2 and 3). Importantly, the addition of ADA3 strongly enhanced the acetylation of histones within chromatin by GCN5L-ADA2b (Fig. (Fig.6B,6B, lane 4 versus lane 3) but not by GCN5L alone (Fig. (Fig.6B,6B, lane 8). The efficient acetylation of chromatin by GCN5L therefore requires ADA2b and ADA3, and this heterotrimer presumably forms a catalytic core in the STAGA complex similar to the one proposed previously for yGcn5-yAda2-yAda3 in SAGA (3). Thus, the assembly of nucleosomes into an array seems to add another barrier to histone tail accessibility to GCN5L compared to the accessibility in mononuclesomes, a barrier that can be overcome with the help of ADA3.

ADA3 deletion mutant proteins ADA3(301-432) and ADA3(109-432) showed little or no enhancement of chromatin acetylation by GCN5L-ADA2b (Fig. (Fig.6B,6B, compare lanes 4, 6, and 7), despite being able to bind ADA2b (see above). The ability to interact with ADA2b is therefore not sufficient for optimal ADA3 function. However, the amino-terminal 50 amino acids of ADA3 seemed to be dispensable for its function as a GCN5L cofactor in this assay, with ADA3(51-432) increasing acetylation even more than the full-length protein (Fig. (Fig.6B6B).

Although we earlier had seen that neither the GCN5L-ADA2a heterodimer nor an ADA2a-containing complex purified from cells could efficiently acetylate mononucleosomes, we further tested whether a GCN5L-ADA2a-ADA3 heterotrimer was able to acetylate chromatin. As can be seen in Fig. Fig.7B,7B, ADA2a was unable to substitute for ADA2b in the efficient acetylation of chromatin by a GCN5L-ADA2-ADA3 heterotrimer (lane 1 versus lanes 7 and 8).

FIG. 7.
The SWIRM domain is dispensable for mononucleosome and chromatin acetylation, but point mutation of the SWIRM domain can lead to decreased acetylation of mononucleosomes. (A) Acetylation of mononuclesomes and free core histones. Substrates were incubated ...

The ADA2b SWIRM domain has a potential role in the regulation of histone acetylation.

As mentioned earlier, both ADA2a and ADA2b have a SWIRM domain that is found in several proteins implicated in chromatin function (2). Figure Figure7E7E shows a comparison of the SWIRM domains of six human proteins. Although similar, the SWIRM domains of ADA2a and ADA2b are less related to each other than the SWIRM domains of the Swi3 homologues BAF155 and BAF170 in the SWI/SNF complex.

Unlike the yAda2 SANT domain, which was found previously to play an important role in histone acetylation by the SAGA complex (7, 50), the SWIRM domain in the metazoan ADA2 homologues has an unknown role. While this work was under way, the structures of several SWIRM domains were published (13, 16, 45, 49, 52) and the SWIRM domain was reported to be a DNA and histone tail binding module. Thus, the SWIRM domain of yeast Swi3 binds DNA and mononucleosomes (16), while the SWIRM domain of ADA2a also binds to DNA (preferentially double stranded) but not to mononucleosomes. The ADA2a SWIRM domain does bind dinucleosomes, however, leading to speculation that it may recognize linker DNA (45). The SWIRM domain of LSD1, on the other hand, was reported to bind the histone tail peptide comprising residues 1 through 20 of H3 (52).

To investigate whether the SWIRM domain of ADA2b plays a role in nucleosome acetylation by STAGA, we purified GCN5L heterodimers with FLAG-tagged ADA2b or ADA2b(1-345), a mutant protein lacking the SWIRM domain (Fig. (Fig.7C,7C, left panel). The two heterodimers and FLAG-tagged GCN5L were normalized based on the amounts of GCN5 (Fig. (Fig.7C,7C, right panel) and compared for their abilities to acetylate free core histones, mononucleosomes, and chromatin. As expected, neither of the two ADA2b proteins affected free histone acetylation by GCN5L (Fig. (Fig.7A,7A, lanes 8 to 11). The SWIRM domain was also dispensable for the ADA2b-dependent enhancement of GCN5L-mediated acetylation of mononucleosomes (Fig. (Fig.7A,7A, lanes 1 to 7) and chromatin (Fig. (Fig.7B,7B, lanes 7 to 9) and for the acetylation of ADA3 (Fig. (Fig.7A,7A, lanes 6 and 7). The efficient acetylation of chromatin and ADA3 in the absence of the ADA2b SWIRM domain suggests that the formation of a heterotrimer is not affected by the absence of the SWIRM domain, in agreement with the results of the in vitro interaction studies.

The SWIRM domain consists of a helical bundle containing a long central helix that separates two smaller helix-turn-helix motifs. The carboxy-terminal fifth helix is thought to bind DNA, since mutation in this domain inhibits DNA or nucleosome binding (16, 45). In particular, the ADA2a amino acid substitution K426A or R428A severely decreases dinucleosome binding, and K429A reduces the affinity of the SWIRM domain for dinucleosomes by at least 50% compared to that of the wild type (45). Intriguingly, however, an alignment of SWIRM domain protein sequences shows that the positively charged K426 of ADA2a is not conserved in other human SWIRM domains, including that of ADA2b, while only ADA2b has a positively charged residue (K403) corresponding to ADA2a R428. However, a positively charged residue corresponding to ADA2a K429 is conserved among all human SWIRM domains (Fig. (Fig.7E7E).

Based on these considerations, we individually mutated ADA2b K403 and R404 (analogous to ADA2a R428 and K429) to alanine and investigated possible effects on histone acetylation by GCN5L. Tagged ADA2b mutant proteins were copurified with untagged GCN5L, which was overexpressed with respect to ADA2b. Mutations of ADA2b K403 and R404 did not influence the interaction of ADA2b with GCN5L (data not shown). The dimers (normalized on the basis of the GCN5 contents) were then tested for their abilities to acetylate either free core histones or mononucleosomes (Fig. (Fig.7D).7D). Unexpectedly, given the results of the deletion mutant analysis presented in Fig. Fig.6A,6A, the R404A SWIRM domain mutation severely impaired the ADA2b-dependent acetylation of mononucleosomes by GCN5 (Fig. (Fig.7D,7D, compare lanes 2 and 6 with lanes 1 and 5) whereas the K403A mutation did not (Fig. (Fig.7D,7D, compare lanes 3 and 7 with lanes 1 and 5). These results suggest a conditional role of the ADA2b SWIRM domain in chromatin acetylation by STAGA. The R404A mutation in the carboxy-terminal fifth helix of the ADA2b SWIRM domain may, as proposed previously (45), disrupt an interaction with the nucleosome that is required when the SWIRM domain is present or lead to a conformational change in ADA2b that blocks the access of other ADA2b regions to the nucleosome.

ADA2a and ADA2b are differentially recruited to p53-activated promoters.

We recently showed that p53-dependent transcription involves the recruitment of GCN5 to target promoters and that p53 binds to STAGA through multivalent interactions with the complex (1, 22). To confirm the recruitment of STAGA by p53 and to determine whether an ADA2a-containing complex distinct from STAGA is tethered to the promoter, we used ChIP assays to assess ADA2a and ADA2b recruitment. Because our antibodies, though paralogue specific, were unsuitable for ChIP assays, we employed the U2OS cells that stably express FLAG-HA-tagged ADA2a or ADA2b. U2OS cells are p53-positive fibroblasts that show induced p21 and GADD45 expression after UV damage (22). An immunoblot of equal amounts of whole-cell extracts from untreated and UV-irradiated U2OS-derived cell lines showed that the tagged-protein levels in the ADA2b- and ADA2a-expressing cell lines were comparable (Fig. (Fig.8A).8A). UV irradiation did not affect the levels of GCN5 or the ADA2 paralogues but did lead to a previously described loss of the unphosphoryated IIa form of RNA polymerase II, as seen by probing with the N20 antibody (41).

FIG. 8.
ADA2a and ADA2b are differentially recruited to activated p53 target gene promoters. (A) Immunoblot of U2OS-derived cells that stably express FLAG-HA-tagged (fha) ADA2a and ADA2b. Equal amounts of whole-cell extracts were resolved by SDS-PAGE, blotted, ...

The GADD45a gene contains a single p53 binding site 1.6 kb downstream of the core promoter in the third intron (32). The human p21 promoter contains two p53 response elements that are located 1.4 and 2.3 kb upstream of the transcription initiation site (19) (Fig. (Fig.8B).8B). Equal amounts of soluble chromatin from UV-irradiated and untreated cells were used for ChIP assays with HA antibody (Fig. 8C and D). The HA antibody coprecipitated tagged ADA2b with the p53 response element of the p21 and GADD45 promoters but not a control DNA fragment. No ADA2a recruitment to either the promoter or the control site was detected (Fig. (Fig.8C).8C). We also used quantitative PCR to evaluate the results of our ChIP experiments (Fig. (Fig.8D).8D). In this analysis, ADA2b, but not ADA2a, was recruited to the p53 response elements of the p21, GADD45, PUMA, and Bax promoters. Thus, unlike the ADA2b-containing STAGA, the yet-to-be-defined human ADA2a-containing complex(es) is not recruited to these p53-dependent promoters under the conditions (UV irradiation) employed here.

DISCUSSION

Like many other histone-modifying factors, STAGA is a large multisubunit complex. The multitude of subunits in these complexes (up to 20 in the case of STAGA) raises the question of which subunits serve purely structural purposes, which play a role in catalysis, and which serve other functions, such as promoter recruitment, through transient interactions with other proteins. Regarding the latter, we recently showed a role for STAGA subunits TAF9, GCN5, and ADA2b in the p53-dependent recruitment of STAGA to target genes (22), and another study has implicated an SPT-TAF module within STAGA in the recruitment of the Mediator complex to MYC target genes (38). Here, we have identified a STAGA catalytic core, composed of GCN5L, ADA2b, and ADA3, and established substrate context-dependent effects of ADA2b and ADA3 on histone acetylation by STAGA. We have further documented nonredundant roles for ADA2b and its paralogue ADA2a, which is not found in STAGA, by demonstrating the selective recruitment of ADA2b to p53 target genes in response to DNA damage.

ADA2a and ADA2b complexes.

Biochemical data indicate that Drosophila homologues of yAda2, Drosophila ADA2a (dADA2a) and dADA2b, are present in the distinct Drosophila GCN5 (dGCN5)-containing ATAC and Drosophila SAGA (dSAGA) complexes, respectively (26, 34). Here, we have shown that the related human homologues ADA2a and ADA2b are also found in distinct GCN5-containing complexes, with ADA2b being specific for STAGA. ADA2a-containing complexes differ from STAGA both in size and in the absence of several previously identified STAGA subunits (40) that include TRRAP and certain SPT or TAF proteins. Unlike the ADA2b-containing STAGA complex, the less well characterized human ADA2a-containing complexes, despite the presence of GCN5 and ADA3 in at least some cases, are not able to acetylate nucleosomes efficiently. These results are in agreement with Drosophila genetic data indicating greatly reduced acetylation of the bona fide dSAGA substrates H3 K9 and H3 K14 in dADA2b-deficient cells but not in cells lacking dADA2a (43). Using recombinant GCN5-ADA2a/ADA2b and GCN5-ADA2a/ADA2b-ADA3 complexes, we have demonstrated that the lack of nucleosomal HAT activity of ADA2a-GCN5-containing complexes is due to the inability of ADA2a, in contrast to ADA2b, to stimulate nucleosomal histone acetylation by GCN5L and is not a result of the inhibition of GCN5 by ADA2a-associated proteins.

The dADA2a- and dGCN5-containing Drosophila ATAC complex is able to acetylate nucleosomal H4 in vitro (26), and genetic data indicate that the loss of dADA2a, but not of dADA2b, leads to decreases in H4 K5 and K12 acetylation in vivo (14). Interestingly, dGCN5 null mutants display similar reductions in H4 acetylation (14). This phenotype can be reversed by a dGCN5 deletion mutant protein lacking the amino-terminal region conserved in human GCN5L and PCAF (a deletion resulting in a protein similar to human GCN5S) but not by dGCN5 containing mutations in the HAT domain or the ADA interaction region (14), leading to speculation that dADA2a and/or associated proteins change the substrate preference of dGCN5 from H3 to H4. Although we failed to see nucleosomal H4 acetylation by an immunoprecipitated ADA2a complex, our data obtained with a recombinant GCN5-ADA2a-ADA3 heterotrimer, which failed to show acetylation of nucleosomal arrays, favor the explanation by Guelman et al. (26) that the dADA2a-containing ATAC complex includes an H4 acetyltransferase distinct from dGCN5, recently identified as CG10414/Atac2 (51).

Histone context-dependent functions of components of the three-subunit catalytic core of STAGA.

The demonstration with yeast proteins that free histones can be acetylated both by purified yGcn5 and by the SAGA complex and that nucleosomes can be acetylated by SAGA but only poorly by yGcn5 alone indicated roles for other SAGA subunits in modulating yGcn5 function (23). The findings of later studies showed that the acetylation of free histones by yGcn5 is enhanced by yAda2, with no additional effect of yAda3, and that the acetylation of nucleosomal histones by yGcn5 is slightly enhanced by yAda2 and markedly enhanced by yAda2 and yAda3. yGcn5, yAda2, and yAda3 have thus been regarded as the catalytic core of SAGA (3). Despite the important information provided by these studies, the significant structural differences between yGcn5 and human GCN5L, as well as the existence of two human homologues of yGcn5 and yAda2, have called for similar studies of human GCN5L, the ADA2a and ADA2b paralogues, and human ADA3.

In the present study, free histones (H3 and H4) were found to be acetylated equally well by isolated GCN5L and by the STAGA complex. This result is consistent with the observed lack of any effect of ADA2a, ADA2b, or ADA3 on isolated GCN5L activity in this assay but contrasts with the reported activity of yAda2 in enhancing free histone acetylation by yGcn5 (3, 7), perhaps reflecting a SANT domain-independent mechanism in the case of the human GCN5 catalysis (see below). The efficient acetylation of histones within mononucleosomes by GCN5L was found to be dependent upon ADA2b but does not require ADA3. In contrast, the efficient acetylation of histones within nucleosomal arrays (chromatin) by GCN5 was found to be dependent upon both ADA2b and ADA3, which were found to form a trimeric complex with GCN5. Thus, in the case of GCN5, the higher-order arrangements of histones, first within mononucleosomes and second within nucleosomal arrays, present at least two levels of constraints to GCN5-mediated histone acetylation that are countered, respectively, by ADA2b and ADA3.

Interestingly, ADA3 itself is a substrate for GCN5 within the heterotrimer, and the autoacetylation of a STAGA subunit of the same size as ADA3 suggests that ADA3 acetylation also occurs within the natural complex. The autoacetylation of another HAT, p300, is associated with catalytic activation but was also proposed previously to enhance the association of p300 with the transcription factor ATF-2 (31) and, perhaps more importantly, to facilitate both the release of p300 following histone acetylation and subsequent preinitiation complex formation (6). We presently do not know the role of STAGA autoacetylation, but a similar role(s) for ADA3 acetylation is conceivable.

Role of ADA2b domains in chromatin acetylation by GCN5L.

The mechanism by which ADA2b and ADA3 increase GCN5L activity on chromatin remains to be elucidated. Plausible mechanistic explanations are a conformational change induced in GCN5L upon heterotrimer formation and additional contacts of the ADA2b and ADA3 molecules with the substrate. In this regard, based on the findings of mutational studies of yGcn5, it was proposed previously that the yGcn5 catalytic domain must undergo a conformational change in the substrate binding groove for efficient binding of histone tails (35), an isomerization that may be influenced by yAda2 and yAda3. An alternative model suggests that the SANT domain of yAda2 presents the substrate to yGcn5 (7).

The results of the present study reveal that human GCN5L, unlike yGcn5, does not require an ADA2 SANT domain for the optimal acetylation of free histones, suggesting that GCN5L has an intrinsic ability to efficiently bind histone tails. However, GCN5L does require ADA2b for the optimal acetylation of nucleosomal histones. It may well be that the ADA2b SANT domain plays a role in the recognition of nucleosomes, in which case it would be interesting to investigate whether/why the related ADA2a SANT domain is unable to substitute for it.

The SWIRM domain is another domain common to yAda2 and its human paralogues. Recent publications have implied a role for SWIRM domains in chromatin binding (16, 45, 52), an intriguing idea given that proteins with SWIRM domains are generally involved in chromatin modifications. Human proteins with SWIRM domains include the herein-described ADA2b and ADA2a paralogues associated with the acetyltansferase GCN5, the yeast Swi3 homologues BAF155 and BAF170 found in the chromatin-remodeling complex SWI/SNF (55), the histone demethylase LSD1/BHC110 (also found in histone deacetylase complexes) (27, 47), and the histone H2A deubiquitinase KIAA1915/MYSM1 (64) (Fig. (Fig.7E).7E). Yet it is already clear that not all SWIRM domains share the ability to bind to DNA (61), indicating that their abilities to bind chromatin may also differ.

The results presented herein indicate that whereas the deletion of the ADA2b SWIRM domain does not abolish the observed stimulatory effect of ADA2b on chromatin acetylation by GCN5L, other SWIRM domain mutations do inhibit that ADA2b function. Thus, the SWIRM domain seems to have a participatory role in chromatin acetylation yet is conditionally required under the assay conditions used in this study. Because linker histone H1 has been shown previously to repress the acetylation of nucleosomes by the GCN5L-related PCAF enzyme and the corresponding PCAF complex (28), and because of suggestions that the SWIRM domain of ADA2a may help overcome H1-mediated repression of chromatin remodeling by ACF (45), we further tested the acetylation of mononucleosomes by GCN5L-ADA2b in the presence and absence of H1. H1 was found to completely repress the stimulatory effect of ADA2b on the acetylation of mononuclesomes by GCN5L, suggesting that the SWIRM domain was not able to overcome the H1-imposed barrier. The results of assays with oligonucleosomes as substrates have shown that, under conditions of low H1/H4 ratios, the PCAF complex, unlike the isolated PCAF enzyme, can overcome the H1-mediated repression of acetylation (28). Similar assays may reveal restraints that impose the requirement for the SWIRM domain of ADA2b. Another approach to further dissect the roles of the SANT and SWIRM domains and to provide insight into the mechanism of chromatin acetylation by GCN5-ADA2b-ADA3 may entail the use of chimeric ADA2a and ADA2b proteins.

ADA3 and chromatin-modifying complexes.

The requirement of ADA3 for the acetylation of histones within chromatin, but not within mononucleosomes, by GCN5 shows that the assembly of nucleosomes into arrays establishes a further barrier against substrate accessibility to histone-modifying enzymes. Unlike ADA2b, ADA3 has no recognizable domain common to other proteins. The results of our studies show that the ADA3 carboxy terminus is important for the interaction with ADA2b and that the deletion of amino acids 51 to 108 severely compromises ADA3 function, suggesting that the latter region, containing conserved patches rich in charged residues, may play an important role in chromatin recognition. However, immunoblots of size-fractionated immunoprecipitates from cells that stably express FLAG-tagged ADA3 have suggested that ADA3 is a subunit of several complexes. In a potentially related finding, the deletion of ADA3 in Drosophila leads to a number of changes in chromatin that include reductions in H3 K9 and K14 acetylation (mediated presumably through dSAGA), H4 K12 acetylation, and H3 S10 phosphorylation. H4 K5 acetylation, which is reduced in dGCN5 and dADA2a mutants, is not affected by dADA3 mutation (24), indicating multiple and complex roles of both ADA3 and GCN5. Further studies on the function of human ADA3, based also on the composition of the various human ADA3-containing complexes, are warranted to provide further insight on the mechanism of chromatin recognition and modification.

Role of the STAGA complex in p53 function.

The transcription factor p53 has been shown previously to recruit a variety of histone-modifying enzymes, including p300/CBP, PRMT1, and CARM1, to p53-dependent genes, resulting in targeted acetylation and the methylation of histones (1, 5, 20). TRRAP, a subunit of the STAGA (40), PCAF (54), TATA binding protein-free TAF-containing (9), TIP60 (30), and p400 (21) HAT complexes, was also found to accumulate at p53-dependent promoters after DNA damage (5). Consistent with these observations, TIP60 (reviewed in reference 44) and GCN5 (1, 22) have been implicated previously in p53 transactivation. In the latter case, it was shown that GCN5 is recruited to p53-dependent promoters, that GCN5 knockdown impairs p53-dependent transcription, and that STAGA directly binds to p53 (1, 22). While these data are consistent with a role for STAGA in p53 transactivation, this role was not proven because of the observation reported herein that GCN5 is a subunit of several complexes. Indeed, we observed that heterotrimers containing GCN5, ADA3, and either ADA2b or ADA2a are able to bind GST-p53 in vitro (data not shown). Nonetheless, our observation that the STAGA-specific ADA2b protein, and not ADA2a, is selectively recruited to p53-dependent promoters after DNA damage in vivo does provide conclusive evidence for the role of STAGA in p53-dependent transcription. In further support of this conclusion, the knockdown of ADA2a did not influence the transcription of the p53-dependent p21 gene (see Fig. S4 in the supplemental material) and the STAGA-associated ubiquitin hydrolase USP22 was found to be important for the induction of several, though not all, p53-dependent genes (62). The abilities of STAGA, in contrast to the ADA2a-containing complex, to acetylate nucleosomal H3, to interact with Mediator (38), to associate with USP22 via TAF5L (63), and possibly to play a role in preinitiation complex formation via SPT3, as has been proposed previously for yeast (18, 48), may account for the differential requirements of the two ADA2 variants in gene activation by p53.

Supplementary Material

[Supplemental material]

Acknowledgments

We thank Yoshihiro Nakatani, Laszlo Tora, and Christina Hughes for providing reagents.

This work was supported in part by NIH grants CA129325, CA113822, and DK071900. A.M.G. was a Burroughs Wellcome Fellow.

Footnotes

[down-pointing small open triangle]Published ahead of print on 20 October 2008.

Supplemental material for this article may be found at http://mcb.asm.org/.

REFERENCES

1. An, W., J. Kim, and R. G. Roeder. 2004. Ordered cooperative functions of PRMT1, p300, and CARM1 in transcriptional activation by p53. Cell 117735-748. [PubMed]
2. Aravind, L., and L. M. Iyer. 2002. The SWIRM domain: a conserved module found in chromosomal proteins points to novel chromatin-modifying activities. Genome Biol. 3RESEARCH0039. [PMC free article] [PubMed]
3. Balasubramanian, R., M. G. Pray-Grant, W. Selleck, P. A. Grant, and S. Tan. 2002. Role of the Ada2 and Ada3 transcriptional coactivators in histone acetylation. J. Biol. Chem. 2777989-7995. [PubMed]
4. Barlev, N. A., A. V. Emelyanov, P. Castagnino, P. Zegerman, A. J. Bannister, M. A. Sepulveda, F. Robert, L. Tora, T. Kouzarides, B. K. Birshtein, and S. L. Berger. 2003. A novel human Ada2 homologue functions with Gcn5 or Brg1 to coactivate transcription. Mol. Cell. Biol. 236944-6957. [PMC free article] [PubMed]
5. Barlev, N. A., L. Liu, N. H. Chehab, K. Mansfield, K. G. Harris, T. D. Halazonetis, and S. L. Berger. 2001. Acetylation of p53 activates transcription through recruitment of coactivators/histone acetyltransferases. Mol. Cell 81243-1254. [PubMed]
6. Black, J. C., J. E. Choi, S. R. Lombardo, and M. Carey. 2006. A mechanism for coordinating chromatin modification and preinitiation complex assembly. Mol. Cell 23809-818. [PubMed]
7. Boyer, L. A., M. R. Langer, K. A. Crowley, S. Tan, J. M. Denu, and C. L. Peterson. 2002. Essential role for the SANT domain in the functioning of multiple chromatin remodeling enzymes. Mol. Cell 10935-942. [PubMed]
8. Brand, M., J. G. Moggs, M. Oulad-Abdelghani, F. Lejeune, F. J. Dilworth, J. Stevenin, G. Almouzni, and L. Tora. 2001. UV-damaged DNA-binding protein in the TFTC complex links DNA damage recognition to nucleosome acetylation. EMBO J. 203187-3196. [PMC free article] [PubMed]
9. Brand, M., K. Yamamoto, A. Staub, and L. Tora. 1999. Identification of TATA-binding protein-free TAFII-containing complex subunits suggests a role in nucleosome acetylation and signal transduction. J. Biol. Chem. 27418285-18289. [PubMed]
10. Bu, P., Y. A. Evrard, G. Lozano, and S. Y. Dent. 2007. Loss of Gcn5 acetyltransferase activity leads to neural tube closure defects and exencephaly in mouse embryos. Mol. Cell. Biol. 273405-3416. [PMC free article] [PubMed]
11. Candau, R., and S. L. Berger. 1996. Structural and functional analysis of yeast putative adaptors. Evidence for an adaptor complex in vivo. J. Biol. Chem. 2715237-5245. [PubMed]
12. Candau, R., P. A. Moore, L. Wang, N. Barlev, C. Y. Ying, C. A. Rosen, and S. L. Berger. 1996. Identification of human proteins functionally conserved with the yeast putative adaptors ADA2 and GCN5. Mol. Cell. Biol. 16593-602. [PMC free article] [PubMed]
13. Chen, Y., Y. Yang, F. Wang, K. Wan, K. Yamane, Y. Zhang, and M. Lei. 2006. Crystal structure of human histone lysine-specific demethylase 1 (LSD1). Proc. Natl. Acad. Sci. USA 10313956-13961. [PMC free article] [PubMed]
14. Ciurciu, A., O. Komonyi, T. Pankotai, and I. M. Boros. 2006. The Drosophila histone acetyltransferase Gcn5 and transcriptional adaptor Ada2a are involved in nucleosomal histone H4 acetylation. Mol. Cell. Biol. 269413-9423. [PMC free article] [PubMed]
15. Cote, J., R. T. Utley, and J. L. Workman. 1995. Basic analysis of transcription factor binding to nucleosomes. Methods Mol. Genet. 6108-120.
16. Da, G., J. Lenkart, K. Zhao, R. Shiekhattar, B. R. Cairns, and R. Marmorstein. 2006. Structure and function of the SWIRM domain, a conserved protein module found in chromatin regulatory complexes. Proc. Natl. Acad. Sci. USA 1032057-2062. [PMC free article] [PubMed]
17. Dignam, J. D., R. M. Lebovitz, and R. G. Roeder. 1983. Accurate transcription initiation by RNA polymerase II in a soluble extract from isolated mammalian nuclei. Nucleic Acids Res. 111475-1489. [PMC free article] [PubMed]
18. Dudley, A. M., C. Rougeulle, and F. Winston. 1999. The Spt components of SAGA facilitate TBP binding to a promoter at a post-activator-binding step in vivo. Genes Dev. 132940-2945. [PMC free article] [PubMed]
19. el-Deiry, W. S., T. Tokino, T. Waldman, J. D. Oliner, V. E. Velculescu, M. Burrell, D. E. Hill, E. Healy, J. L. Rees, S. R. Hamilton, et al. 1995. Topological control of p21WAF1/CIP1 expression in normal and neoplastic tissues. Cancer Res. 552910-2919. [PubMed]
20. Espinosa, J. M., and B. M. Emerson. 2001. Transcriptional regulation by p53 through intrinsic DNA/chromatin binding and site-directed cofactor recruitment. Mol. Cell 857-69. [PubMed]
21. Fuchs, M., J. Gerber, R. Drapkin, S. Sif, T. Ikura, V. Ogryzko, W. S. Lane, Y. Nakatani, and D. M. Livingston. 2001. The p400 complex is an essential E1A transformation target. Cell 106297-307. [PubMed]
22. Gamper, A. M., and R. G. Roeder. 2008. Multivalent binding of p53 to the STAGA complex mediates coactivator recruitment after UV damage. Mol. Cell. Biol. 282517-2527. [PMC free article] [PubMed]
23. Grant, P. A., L. Duggan, J. Cote, S. M. Roberts, J. E. Brownell, R. Candau, R. Ohba, T. Owen-Hughes, C. D. Allis, F. Winston, S. L. Berger, and J. L. Workman. 1997. Yeast Gcn5 functions in two multisubunit complexes to acetylate nucleosomal histones: characterization of an Ada complex and the SAGA (Spt/Ada) complex. Genes Dev. 11:1640-1650. [PubMed]
24. Grau, B., C. Popescu, L. Torroja, D. Ortuno-Sahagun, I. Boros, and A. Ferrus. 2008. Transcriptional adaptor ADA3 of Drosophila melanogaster is required for histone modification, position effect variegation, and transcription. Mol. Cell. Biol. 28376-385. [PMC free article] [PubMed]
25. Gu, W., and R. G. Roeder. 1997. Activation of p53 sequence-specific DNA binding by acetylation of the p53 C-terminal domain. Cell 90595-606. [PubMed]
26. Guelman, S., T. Suganuma, L. Florens, S. K. Swanson, C. L. Kiesecker, T. Kusch, S. Anderson, J. R. Yates III, M. P. Washburn, S. M. Abmayr, and J. L. Workman. 2006. Host cell factor and an uncharacterized SANT domain protein are stable components of ATAC, a novel dAda2A/dGcn5-containing histone acetyltransferase complex in Drosophila. Mol. Cell. Biol. 26871-882. [PMC free article] [PubMed]
27. Hakimi, M. A., Y. Dong, W. S. Lane, D. W. Speicher, and R. Shiekhattar. 2003. A candidate X-linked mental retardation gene is a component of a new family of histone deacetylase-containing complexes. J. Biol. Chem. 2787234-7239. [PubMed]
28. Herrera, J. E., K. L. West, R. L. Schiltz, Y. Nakatani, and M. Bustin. 2000. Histone H1 is a specific repressor of core histone acetylation in chromatin. Mol. Cell. Biol. 20523-529. [PMC free article] [PubMed]
29. Horiuchi, J., N. Silverman, G. A. Marcus, and L. Guarente. 1995. ADA3, a putative transcriptional adaptor, consists of two separable domains and interacts with ADA2 and GCN5 in a trimeric complex. Mol. Cell. Biol. 151203-1209. [PMC free article] [PubMed]
30. Ikura, T., V. V. Ogryzko, M. Grigoriev, R. Groisman, J. Wang, M. Horikoshi, R. Scully, J. Qin, and Y. Nakatani. 2000. Involvement of the TIP60 histone acetylase complex in DNA repair and apoptosis. Cell 102463-473. [PubMed]
31. Karanam, B., L. Wang, D. Wang, X. Liu, R. Marmorstein, R. Cotter, and P. A. Cole. 2007. Multiple roles for acetylation in the interaction of p300 HAT with ATF-2. Biochemistry 468207-8216. [PMC free article] [PubMed]
32. Kastan, M. B., Q. Zhan, W. S. el-Deiry, F. Carrier, T. Jacks, W. V. Walsh, B. S. Plunkett, B. Vogelstein, and A. J. Fornace, Jr. 1992. A mammalian cell cycle checkpoint pathway utilizing p53 and GADD45 is defective in ataxia-telangiectasia. Cell 71:587-597. [PubMed]
33. Kouzarides, T. 2007. Chromatin modifications and their function. Cell 128693-705. [PubMed]
34. Kusch, T., S. Guelman, S. M. Abmayr, and J. L. Workman. 2003. Two Drosophila Ada2 homologues function in different multiprotein complexes. Mol. Cell. Biol. 233305-3319. [PMC free article] [PubMed]
35. Langer, M. R., K. G. Tanner, and J. M. Denu. 2001. Mutational analysis of conserved residues in the GCN5 family of histone acetyltransferases. J. Biol. Chem. 27631321-31331. [PubMed]
36. Lee, K. K., and J. L. Workman. 2007. Histone acetyltransferase complexes: one size doesn't fit all. Nat. Rev. Mol. Cell Biol. 8284-295. [PubMed]
37. Li, B., M. Carey, and J. L. Workman. 2007. The role of chromatin during transcription. Cell 128707-719. [PubMed]
38. Liu, X., M. Vorontchikhina, Y. L. Wang, F. Faiola, and E. Martinez. 2008. STAGA recruits Mediator to the MYC oncoprotein to stimulate transcription and cell proliferation. Mol. Cell. Biol. 28108-121. [PMC free article] [PubMed]
39. Martinez, E., T. K. Kundu, J. Fu, and R. G. Roeder. 1998. A human SPT3-TAFII31-GCN5-L acetylase complex distinct from transcription factor IID. J. Biol. Chem. 27323781-23785. [PubMed]
40. Martinez, E., V. B. Palhan, A. Tjernberg, E. S. Lymar, A. M. Gamper, T. K. Kundu, B. T. Chait, and R. G. Roeder. 2001. Human STAGA complex is a chromatin-acetylating transcription coactivator that interacts with pre-mRNA splicing and DNA damage-binding factors in vivo. Mol. Cell. Biol. 216782-6795. [PMC free article] [PubMed]
41. McKay, B. C., F. Chen, S. T. Clarke, H. E. Wiggin, L. M. Harley, and M. Ljungman. 2001. UV light-induced degradation of RNA polymerase II is dependent on the Cockayne's syndrome A and B proteins but not p53 or MLH1. Mutat. Res. 48593-105. [PubMed]
42. Ogryzko, V. V., T. Kotani, X. Zhang, R. L. Schiltz, T. Howard, X. J. Yang, B. H. Howard, J. Qin, and Y. Nakatani. 1998. Histone-like TAFs within the PCAF histone acetylase complex. Cell 9435-44. [PubMed]
43. Pankotai, T., O. Komonyi, L. Bodai, Z. Ujfaludi, S. Muratoglu, A. Ciurciu, L. Tora, J. Szabad, and I. Boros. 2005. The homologous Drosophila transcriptional adaptors ADA2a and ADA2b are both required for normal development but have different functions. Mol. Cell. Biol. 258215-8227. [PMC free article] [PubMed]
44. Park, J. H., and R. G. Roeder. 2006. GAS41 is required for repression of the p53 tumor suppressor pathway during normal cellular proliferation. Mol. Cell. Biol. 264006-4016. [PMC free article] [PubMed]
45. Qian, C., Q. Zhang, S. Li, L. Zeng, M. J. Walsh, and M. M. Zhou. 2005. Structure and chromosomal DNA binding of the SWIRM domain. Nat. Struct. Mol. Biol. 121078-1085. [PubMed]
46. Sadowski, H. B., and M. Z. Gilman. 1993. Cell-free activation of a DNA-binding protein by epidermal growth factor. Nature 36279-83. [PubMed]
47. Shi, Y., F. Lan, C. Matson, P. Mulligan, J. R. Whetstine, P. A. Cole, R. A. Casero, and Y. Shi. 2004. Histone demethylation mediated by the nuclear amine oxidase homolog LSD1. Cell 119941-953. [PubMed]
48. Shukla, A., P. Bajwa, and S. R. Bhaumik. 2006. SAGA-associated Sgf73p facilitates formation of the preinitiation complex assembly at the promoters either in a HAT-dependent or independent manner in vivo. Nucleic Acids Res. 346225-6232. [PMC free article] [PubMed]
49. Stavropoulos, P., G. Blobel, and A. Hoelz. 2006. Crystal structure and mechanism of human lysine-specific demethylase-1. Nat. Struct. Mol. Biol. 13626-632. [PubMed]
50. Sterner, D. E., X. Wang, M. H. Bloom, G. M. Simon, and S. L. Berger. 2002. The SANT domain of Ada2 is required for normal acetylation of histones by the yeast SAGA complex. J. Biol. Chem. 2778178-8186. [PubMed]
51. Suganuma, T., J. L. Gutierrez, B. Li, L. Florens, S. K. Swanson, M. P. Washburn, S. M. Abmayr, and J. L. Workman. 2008. ATAC is a double histone acetyltransferase complex that stimulates nucleosome sliding. Nat. Struct. Mol. Biol. 15364-372. [PubMed]
52. Tochio, N., T. Umehara, S. Koshiba, M. Inoue, T. Yabuki, M. Aoki, E. Seki, S. Watanabe, Y. Tomo, M. Hanada, M. Ikari, M. Sato, T. Terada, T. Nagase, O. Ohara, M. Shirouzu, A. Tanaka, T. Kigawa, and S. Yokoyama. 2006. Solution structure of the SWIRM domain of human histone demethylase LSD1. Structure 14457-468. [PubMed]
53. Tremethick, D. J. 2007. Higher-order structures of chromatin: the elusive 30 nm fiber. Cell 128651-654. [PubMed]
54. Vassilev, A., J. Yamauchi, T. Kotani, C. Prives, M. L. Avantaggiati, J. Qin, and Y. Nakatani. 1998. The 400 kDa subunit of the PCAF histone acetylase complex belongs to the ATM superfamily. Mol. Cell 2869-875. [PubMed]
55. Wang, W., Y. Xue, S. Zhou, A. Kuo, B. R. Cairns, and G. R. Crabtree. 1996. Diversity and specialization of mammalian SWI/SNF complexes. Genes Dev. 102117-2130. [PubMed]
56. Wieczorek, E., M. Brand, X. Jacq, and L. Tora. 1998. Function of TAF(II)-containing complex without TBP in transcription by RNA polymerase II. Nature 393187-191. [PubMed]
57. Xu, W., D. G. Edmondson, Y. A. Evrard, M. Wakamiya, R. R. Behringer, and S. Y. Roth. 2000. Loss of Gcn5l2 leads to increased apoptosis and mesodermal defects during mouse development. Nat. Genet. 26229-232. [PubMed]
58. Xu, W., D. G. Edmondson, and S. Y. Roth. 1998. Mammalian GCN5 and P/CAF acetyltransferases have homologous amino-terminal domains important for recognition of nucleosomal substrates. Mol. Cell. Biol. 185659-5669. [PMC free article] [PubMed]
59. Yamauchi, T., J. Yamauchi, T. Kuwata, T. Tamura, T. Yamashita, N. Bae, H. Westphal, K. Ozato, and Y. Nakatani. 2000. Distinct but overlapping roles of histone acetylase PCAF and of the closely related PCAF-B/GCN5 in mouse embryogenesis. Proc. Natl. Acad. Sci. USA 9711303-11306. [PMC free article] [PubMed]
60. Yang, X. J., V. V. Ogryzko, J. Nishikawa, B. H. Howard, and Y. Nakatani. 1996. A p300/CBP-associated factor that competes with the adenoviral oncoprotein E1A. Nature 382319-324. [PubMed]
61. Yoneyama, M., N. Tochio, T. Umehara, S. Koshiba, M. Inoue, T. Yabuki, M. Aoki, E. Seki, T. Matsuda, S. Watanabe, Y. Tomo, Y. Nishimura, T. Harada, T. Terada, M. Shirouzu, Y. Hayashizaki, O. Ohara, A. Tanaka, T. Kigawa, and S. Yokoyama. 2007. Structural and functional differences of SWIRM domain subtypes. J. Mol. Biol. 369222-238. [PubMed]
62. Zhang, X. Y., M. Varthi, S. M. Sykes, C. Phillips, C. Warzecha, W. Zhu, A. Wyce, A. W. Thorne, S. L. Berger, and S. B. McMahon. 2008. The putative cancer stem cell marker USP22 is a subunit of the human SAGA complex required for activated transcription and cell-cycle progression. Mol. Cell 29102-111. [PMC free article] [PubMed]
63. Zhao, Y., G. Lang, S. Ito, J. Bonnet, E. Metzger, S. Sawatsubashi, E. Suzuki, X. Le Guezennec, H. G. Stunnenberg, A. Krasnov, S. G. Georgieva, R. Schule, K. Takeyama, S. Kato, L. Tora, and D. Devys. 2008. A TFTC/STAGA module mediates histone H2A and H2B deubiquitination, coactivates nuclear receptors, and counteracts heterochromatin silencing. Mol. Cell 2992-101. [PubMed]
64. Zhu, P., W. Zhou, J. Wang, J. Puc, K. A. Ohgi, H. Erdjument-Bromage, P. Tempst, C. K. Glass, and M. G. Rosenfeld. 2007. A histone H2A deubiquitinase complex coordinating histone acetylation and H1 dissociation in transcriptional regulation. Mol. Cell 27609-621. [PMC free article] [PubMed]

Articles from Molecular and Cellular Biology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links