A Human SPT3-TAFII31-GCN5-L Acetylase Complex Distinct from Transcription Factor IID*

In yeast, SPT3 is a component of the multiprotein SPT-ADA-GCN5 acetyltransferase (SAGA) complex that integrates proteins with transcription coactivator/adaptor functions (ADAs and GCN5), histone acetyltransferase activity (GCN5), and core promoter-selective functions (SPTs) involving interactions with the TATA-binding protein (TBP). In particular, yeast SPT3 has been shown to interact directly with TBP. Here we report the molecular cloning of a cDNA encoding a human homologue of yeast SPT3. Amino acid sequence comparisons between human SPT3 (hSPT3) and its counterparts in different yeast species reveal three highly conserved domains, with the most conserved 92-amino acid N-terminal domain being 25% identical with human TAFII18. Despite the significant sequence similarity with TAFII18, native hSPT3 is not a bona fide TAFII because it is not associated in vivoeither with human TBP/TFIID or with a TFIID-related TBP-free TAFII complex. However, we present evidence that hSPT3 is associated in vivo with TAFII31 and the recently described longer form of human GCN5 (hGCN5-L) in a novel human complex that has histone acetyltransferase activity. We propose that the human SPT3-TAFII31-GCN5-L acetyltransferase (STAGA) complex is a likely homologue of the yeast SAGA complex.

Yeast SPT (suppressors of Ty) 1 genes, including SPT3, encode global transcription regulators and were originally identified in a genetic screen for mutations that suppress transcriptional defects caused by the insertion of the retrotransposon Ty or its long terminal repeat, ␦, in the promoter region of several genes (for a review see Ref. 1). This approach also identified the gene encoding the yeast TATA-binding protein (TBP) as SPT15 (2,3). However, in contrast to SPT15, SPT3 is not essential for yeast viability. Genetic and biochemical analyses have shown that SPT3 and TBP interact in yeast (4), and mutations in SPT3, SPT7, and SPT8, as well as particular mutations in TBP/SPT15, all result in a common set of phenotypes that include slow growth and defects in mating and sporulation (2,5,6). Accordingly, deletion of the SPT3 gene in yeast results in gene-selective RNA polymerase II transcription defects (5)(6)(7). The mechanisms for the gene-specific functions of SPT3 are still poorly understood but may include core promoter-selective functions of SPT3 in TATA box selection. Indeed, SPT3 has been proposed to facilitate TBP recruitment to weak TATAcontaining or TATA-less promoters in yeast (4,8). Consistent with this notion, TFIIA overexpression in yeast partially suppresses an spt3⌬ mutation, and spt3⌬/toa1 (TFIIA) double mutants are inviable (9). More recently, yeast SPT3 has been shown to be part of the 1.8-MDa multiprotein yeast SAGA (SPT-ADA-GCN5 acetyltransferase) complex that also contains SPT7, SPT8, SPT20/ADA5, and the coactivators/adaptors ADA1, ADA2, ADA3, and GCN5 (10,11). Altogether these observations suggest an important role for SPT3 (as well as SPT7, SPT8, and SPT20) in linking core promoter-specific functions (e.g. stabilization of TBP/TFIID-DNA interactions) in vivo to upstream activators through an adaptor/coactivator complex(es) with histone acetyltransferase activity.
Recently, putative human homologues of components of the yeast SAGA complex have been isolated. These include hADA2 (12) and three human GCN5 acetyltransferase family members: PCAF (p300/CBP-associated factor) (13), a short 55-kDa hGCN5 (hGCN5-S) (12,13), and a long 93-kDa hGCN5 (hGCN5-L) (14). The short and long hGCN5 forms are produced from the same gene, presumably by alternatively spliced mRNAs. The longer hGCN5-L contains a 361-amino acid Nterminal domain (the PCAF homology domain) that is absent in hGCN5-S and yGCN5. This domain shares significant homology with the corresponding 351-amino acid N-terminal domain of PCAF that interacts with the coactivator p300/CBP (13,14). Here we describe the molecular cloning of a cDNA encoding a human homologue of yeast SPT3. We present evidence for a specific association in vivo of human SPT3 with TAF II 31 (TBPassociated factor II 31) and hGCN5-L in a novel human complex that is distinct from TFIID and that has histone acetyltransferase activity with preference for histone H3. Our results together with those just reported by Ogryzko et al. (15) suggest that the human SPT3-TAF II 31-GCN5-L acetyltransferase (STAGA) complex is one of perhaps several distinct human homologues of the yeast SAGA complex.

EXPERIMENTAL PROCEDURES
Molecular Cloning of Human SPT3 cDNA-A search of the Gen-Bank TM EST division with the yeast SPT3 sequence revealed a human EST sequence (N89343) 36% identical with yeast SPT3 amino acids 7-47 and a mouse EST sequence (W71809) 42% identical with yeast SPT3 amino acids 44 -88. A human SPT3 (hSPT3) cDNA fragment was obtained from a Marathon-Ready HeLa cDNA library (CLONTECH) by nested PCR using degenerate primers in the mouse 3Ј-end EST sequence and primers in the human 5Ј-end EST sequence. Rapid amplification of cDNA ends and high fidelity PCR with cloned Pfu polymerase (Stratagene) were then used to obtain, from the same library, the full-length hSPT3 cDNA. The sequence was confirmed from at least two independent clones. The hSPT3 cDNA sequence has been deposited in GenBank TM with the accession number AF073930. For efficient expression of full-length recombinant hSPT3 protein in bacteria, hSPT3 cDNA nucleotides 120 -128 (GGA AGG AGT; 3 codons for Gly, Arg, and Ser, respectively) were recoded to GGT CGT TCT (the silent changes are underlined) to remove a fortuitous bacterial ribosome binding site. The recoded hSPT3 cDNA, which also contained a newly created NdeI site at the first methionine and a BamHI site insertion after the natural stop codon at position 1031, was inserted between the NdeI and BamHI sites of 6hisT-pET11d (16) to obtain the bacterial expression vector pET-6His-hSPT3.
Northern Blot Analysis-A human multiple tissue Northern blot (CLONTECH) was probed with 32 P-labeled cDNA probes. The hSPT3 cDNA probe was a PCR fragment from nucleotides 360 -811. Human cDNAs for TAF II 150 2 and ␤-actin (CLONTECH) were used as reference probes on the same blot (after stripping it).
Cell Lines, Nuclear Extracts, Antibodies, and Immunoprecipitations-Human HeLa cell derivatives stably expressing FLAG-tagged human TBP (3-10) (17) and human TAF II 100 2 have been described. The human cell line stably expressing FLAG-tagged human TAF II 135 will be described elsewhere. 3 Nuclear extracts were prepared as described previously (18). Rabbit polyclonal antibodies against hSPT3 (No. 623) were raised (Covance) against a bacterially expressed insoluble recombinant 6-His-tagged hSPT3 protein fragment (amino acids 1-285) that was purified on Ni 2ϩ -NTA-agarose (Qiagen) and by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and excision of the protein band. Rabbit polyclonal antibodies against human TBP (19), TAF II 31 (20), the short form of hGCN5 (13,14), and the N-terminal domain of PCAF (13,14) were described previously. Rabbit polyclonal antibodies against human TAF II 135 will be described elsewhere. 3 Monoclonal anti-FLAG M2 antibody-agarose was from Kodak-IBI. Purification of FLAG epitope-tagged TBP-containing TFIID (eTFIID) from nuclear extracts FIG. 1. Human SPT3 sequence and homology to yeast SPT3 and TAF II 18. A, human SPT3 cDNA and translated amino acid sequence. B, human multiple tissue Northern blot hybridized with radiolabeled cDNA probes for hSPT3, hTAF II 150, and h␤-actin mRNAs. Sk. muscle, skeletal muscle. C, multiple alignments of SPT3 sequences from human (hSPT3), Schizosaccharomyces pombe (S.p.SPT3), S. cerevisiae (S.c.SPT3), Kluyvieromyces lactis (K.l.SPT3), Clavispora opuntiae (C.o.SPT3), and sequences of human TAF II 18 (hTAF18) and its yeast S. cerevisiae homologue (S.c.Fun81). Identical (in bold and dark-shaded) and similar (light-shaded) amino acids that are conserved in at least four sequences are outlined. Brackets above the hSPT3 sequence localize the three highly conserved domains A (N-terminal), B (middle), and C (C-terminal). The arrowhead labeled E3 K identifies the position of the strongest suppressor mutation in S.c.SPT3 of the SPT15-21/TBP mutant phenotype. D, schematic representation and alignment of the three A, B, and C domains in the different molecules with indication of the percentage of amino acid identities shared for each domain between two adjacent molecules. The line ySPT3 symbolizes the consensus sequence for ySPT3 from the different yeast species. The fragments of hTAF II 18 previously shown to interact with hTAF II 28 and hTAF II 30 (see text) are shown under the yFUN81 line. No similarity to other previously cloned proteins was found for the SPT3-specific region containing domains B and C. of the 3-10 cell line was as described previously (17). For immunoprecipitations antibodies were cross-linked to protein A-agarose with dimethylpimelimidate (Sigma). The antibody resin (10 -50 l) was mixed with nuclear extracts (400 -500 l) for 5-12 h at 4°C in binding Buffer C (BC) (20 mM Tris-HCl, pH 7.9, 20% glycerol, 0.2 mM EDTA, 0.05% Nonidet P-40, 8 mM 2-mercaptoethanol, 0.2 mM phenylmethylsulfonyl fluoride) containing 150 mM KCl (BC150) or up to 300 mM KCl (BC300) as indicated. The immune complexes were recovered by low speed centrifugation, and the resin was washed extensively with binding buffer and with BC100 and then eluted with either 20 mM Tris-HCl (pH 8.0) containing 2% SDS or with 0.2 mg/ml FLAG peptide as described previously (17). Western blot analyses were performed by standard procedures and with the ECL detection system (Amersham).
Purification of Human Core Histones and Inositol Phosphate-Histone Acetyltransferase (HAT) Assays-HeLa cell nuclear pellets (18) were used to purify core histones. The nuclear pellet (5 ml) was homogenized with a blender in 40 ml of buffer A (0.1 M potassium phosphate, pH 6.7, 0.1 mM EDTA, 10% glycerol, 0.1 mM phenylmethylsulfonyl fluoride, 0.1 mM dithiothreitol) containing 0.63 M NaCl and centrifuged in a Ti45 rotor (Beckman) at 25,000 rpm at 4°C. The supernatant was mixed and incubated at 4°C with 18 ml of preswollen Bio-Gel-HTP resin (DNA grade, Bio-Rad) for 3 h. The resin was packed into an econo column (Bio-Rad) and washed extensively (0.5 column volume/h, overnight) with buffer A containing 0.63 M NaCl. Core histones were eluted with buffer A containing 2 M NaCl and dialyzed first against buffer B (10 mM potassium phosphate, pH 6.7, 150 mM KCl, 10% glycerol) for 3 h and then against a buffer containing 20 mM Tris-HCl (pH 7.9), 100 mM KCl, 20% glycerol, and 0.1 mM dithiothreitol for 3 h. For the inositol phosphate-HAT assays immunoprecipitations were performed in BC200 as described above, except that BC100 was replaced with HAT assay buffer (50 mM Tris-HCl, pH 8.0, 70 mM KCl, 10% glycerol, 0.1 mM EDTA, 0.05% Nonidet P-40, 10 mM sodium butyrate, 1 mM 2-mercaptoethanol, 0.2 mM phenylmethylsulfonyl fluoride) in the final washes of the immune complexes that were then used directly for the HAT assays. The HAT assays were performed at 30°C for 30 min in HAT assay buffer in a final volume of 25 l and contained 10 -15 l of resinimmune complexes (or either 10 ng of recombinant PCAF or 50 ng of recombinant p300 HAT domain), 1.6 g of purified HeLa core histones, and 125 nCi of [ 3 H]acetyl-CoA (3.8 Ci/mmol, 250 Ci/ml). The reactions were then either analyzed by SDS-PAGE and Coomassie staining followed by fluorography for 16 -24 h at Ϫ70°C or spotted directly onto Whatman P-81 filters that were then washed with 50 mM sodium carbonate buffer (pH 9.2) and counted in a liquid scintillator. Recombinant FLAG-tagged p300 HAT domain (1195-1810) was expressed in bacteria and purified as reported previously (21). Recombinant human PCAF was a kind gift from Y. Nakatani.

RESULTS AND DISCUSSION
Because of the important role of SPT3 in the regulation of TBP/TFIID functions in a core promoter-specific manner in yeast, and because of the core promoter-specific functions of both yeast and human TFIID/TAF II s (for reviews see Refs. [22][23][24], we searched for a potential human homologue of yeast SPT3. A Blast alignment (25) of GenBank TM data base sequences with the yeast (Saccharomyces cerevisiae) SPT3 (ySPT3) protein sequence identified two overlapping mouse and human EST sequences that together encoded a hypothetical protein with significant identity to ySPT3 amino acids 7-88 (see "Experimental Procedures"). This information allowed us to clone by PCR a full-length human cDNA of 1,165 nucleotides, including part of the poly(A) tract, that encodes a 317-amino acid protein with a calculated molecular weight of 35,790 and 27% overall sequence identity to ySPT3 (Fig. 1, A  and C). This suggests that this cDNA encodes a human homologue of ySPT3 that will be referred to hereafter as hSPT3.
A multiple tissue Northern blot analysis revealed a specific 1.4-kilobase hSPT3 mRNA that is approximately the size of the cloned cDNA and is expressed in all human tissues tested in a manner similar to the ubiquitously expressed ␤-actin and hTAF II 150 mRNAs (Fig. 1B). Interestingly, a longer and less abundant 2.5-kilobase mRNA with a more restricted tissue distribution was also detected (Fig. 1B, SPT3 2.5 kilobases), suggesting the possible existence of an additional longer tissuespecific variant of hSPT3.
A Blast alignment of the hSPT3 protein sequence with the protein sequences in the data bases retrieved all the cloned SPT3 genes of various yeast species as well as hTAF II 18 and its S. cerevisiae homologue FUN81/yTAF II 19 (Fig. 1C). The interspecies SPT3 alignments presented in Fig. 1C and schematized in Fig. 1D reveal a high degree of conservation between human and yeast SPT3 in three domains (A, B, and C) that most likely reflects functional conservation. Interestingly, the 92-amino acid domain A of hSPT3 is 38% identical to the yeast SPT3 A domains and 25% identical to human TAF II 18 and its yeast homologue FUN81 (Fig. 1D). This strongly suggests that the SPT3 domain A may fold in a structure similar to TAF II 18 and may have related functions. Since hTAF II 18 has been shown to interact directly with hTAF II 28 and hTAF II 30 (26), the hSPT3 domain A also may serve as a TAF II -interacting surface, possibly also contacting hTAF II 28 and/or hTAF II 30. No homologies with other known proteins in the data bases were found for the less conserved domains B and C, suggesting that these regions may perform SPT3-specific functions. Interestingly, the SPT3 glutamic acid that is mutated to a lysine in the strongest yeast SPT3 suppressor mutant of the spt15-21 (TBP mutant) phenotype (4) is conserved in domain B between human and all yeast SPT3 proteins (Fig. 1C, E3 K). This may indicate a possible function of domain B in direct interactions with TBP.
The above observations suggested that hSPT3 may interact with TFIID through direct contacts with either TBP, as in the case of its yeast counterpart, or TAF II s. We addressed this by testing for the presence of hSPT3 in highly purified human TFIID and by analyzing the physiological interacting partners of human SPT3 in HeLa cells. Highly purified eTFIID was shown to lack any detectable hSPT3 by immunoblot analyses (Fig. 2A, lane 2), whereas a specific 37-kDa hSPT3 protein was detected in the crude HeLa nuclear extract (lane 1). Immunopurification of eTFIID through its FLAG-tagged TBP subunit was performed after two chromatographic steps, including a high salt (0.85 M KCl) elution from phosphocellulose P11. Therefore, it remained possible that the resins or high salt could have disrupted a potential interaction between hSPT3 and TFIID and/or that the FLAG epitope at the N terminus of TBP might have interfered with hSPT3 association with eT-FIID. To further address this issue we performed direct immunoprecipitations both from nuclear extracts of cells expressing FLAG-tagged TAF II 100 (f:TAF II 100) and TAF II 135 (f: TAF II 135) and from nuclear extracts of normal HeLa cells with, respectively, antibodies against the FLAG epitope and against native TAF II 31 and TBP. Anti-TBP (Fig. 2A, lane 6), anti-f:  TAF II 100 (lane 3), and anti-f:TAF II 135 (lane 4) immunoprecipitations all failed to co-precipitate hSPT3, in agreement with the above results, whereas they efficiently precipitated TBP and TAF II s. These results demonstrate that hSPT3 is not a component of the human TFIID complex and that it is unlikely to be part of any other TAF II complex containing TAF II 100 and/or TAF II 135, such as, for instance, the recently described TBP-free TAF II complex TFTC (27). While these results do not exclude direct physical interactions between hSPT3 and TBP/ TFIID, as suggested previously by co-immunoprecipitations of overexpressed ySPT3 and yTBP in yeast cells and by corresponding genetic interactions (4), they do emphasize that in human cells native hSPT3 does not efficiently interact with TBP/TFIID in solution. Thus, a more efficient interaction may require the presence of additional components, such as promoter DNA and associated general transcription factors and/or activators. Interestingly, however, under the same conditions anti-TAF II 31 antibodies efficiently immunoprecipitated hSPT3 in addition to TFIID components ( Fig. 2A, lane 5; Fig. 2B, lane 3), suggesting that in human cells hSPT3 and TAF II 31 are in association in a complex distinct from TFIID (and TFTC). This was further confirmed by immunoprecipitations with anti-hSPT3 antibodies that also coprecipitated TAF II 31 but not TBP (Fig. 2B, lane 4), TAF II 18, TAF II 80, TAF II 100, or TAF II 135 (data not shown).
Because yeast SPT3 is associated with GCN5 histone acetyltransferase in the SAGA complex (10), we tested whether the hSPT3-TAF II 31 complex also has histone acetyltransferase (HAT) activity. Fig. 3A shows that immune complexes obtained with both anti-hSPT3 and anti-TAF II 31 have significant HAT activity when compared with mock (protein A resin alone) or anti-TBP immunoprecipitates. To address the type of HAT involved we compared the pattern of core histone acetylation by the hSPT3-TAF II 31 complex with that of PCAF and p300. The results presented in Fig. 3B indicate that immune complexes obtained with anti-hSPT3 (lane 5) and anti-TAF II 31 (lane 6) both preferentially acetylate histone H3. This suggests that the HAT associated with hSPT3 and TAF II 31 is different from p300, which acetylates all core histones with a preference for H3 and H4 (21) (lane 7), and more related to the human GCN5 family member PCAF (lane 2). In accord with this, and while this manuscript was being prepared, we learned that immunoprecipitations of ectopic FLAG-tagged PCAF and FLAG-tagged hGCN5-S from HeLa cell lines stably overexpressing these HAT factors also coprecipitated hSPT3 and TAF II 31, as well as TAF II 20, TAF II 30, and additional proteins that include novel TAF II -related factors (15). Interestingly, however, we did not find significant amounts of PCAF (Fig. 3C, lane 6 in top panel) or hGCN5-S (lane 6 in bottom panel) in our immunoprecipitated complexes. Instead, we detected predominantly the recently described long form (hGCN5-L) of hGCN5 (Fig. 3C, lane 6 in third panel from the top; and Fig. 3D, lanes 2-5). The reason for the apparent absence of PCAF and hGCN5-S in our human SPT3-TAF II 31-GCN5-L acetylase (STAGA) complex is not clear. However, this most likely results from the different immunoprecipitation approaches used here and in the recent study by Ogryzko et al. (15). One possibility is that our anti-hSPT3 and anti-TAF II 31 antibodies, including different antibodies against the N-terminal and C- and short (hGCN5-S) forms of hGCN5 are indicated. D, Western blot analysis with anti-TAF II 31, anti-hSPT3, anti-TBP, and affinity-purified anti-hGCN5-S antibodies of immune complexes obtained from HeLa cell nuclear extracts with anti-TBP (lane 1), anti-hSPT3 (lane 2), and anti-TAF II 31 (lanes 3-5) antibodies as described above and in Fig. 2A. terminal regions of TAF II 31 (data not shown), all dissociated PCAF and hGCN5-S, but not hGCN5-L, from the STAGA complex(es). Another interesting and more likely possibility is that the PCAF, hGCN5-S, and STAGA complexes are distinct and differ with respect to their associated HAT subunits (and perhaps other components as well) and their relative abundance in HeLa cells. Indeed, this is also suggested by the clear indication that the composition of the PCAF and hGCN5-S complexes are indistinguishable except for the corresponding overexpressed HAT subunits (15). Related to this and in accord with a recent report (14), our results indicate that PCAF is apparently not very abundant in HeLa cells, given the difficulty in detecting it in crude nuclear extracts by immunoblot analyses that efficiently detect the recombinant PCAF protein (Fig. 3C, lane 2 in top panel). It also is important to note that in contrast to the analyses of Ogryzko et al. (15), our immunoprecipitation analyses were performed with antibodies against two different native subunits of the STAGA complex that were not overexpressed. Hence, we propose from these results that hSPT3 and TAF II 31 are predominantly associated with hGCN5-L in HeLa cells.
In conclusion, our finding that human SPT3 exists in a novel in vivo complex (STAGA) with TAF II 31 and the recently described hGCN5-L histone acetyltransferase (14) demonstrates the existence of TAF II s in complexes distinct from TFIID and the recently described TFTC (27). This is in accord with the very recent complementary findings of TAF II s within the yeast SAGA complex (28) and in the human PCAF and hGCN5-S complexes (15). It also suggests a possible diversity of human homologues of the yeast SAGA complex that differ (at least in part) by the associated HAT subunit and perhaps also by their relative abundance/activity in different tissues. This is also supported by the very recent study on PCAF and hGCN5-S complexes (15) and by the observed higher steady state PCAF mRNA levels in muscle as compared with other tissues (13). The future structural and functional characterizations of these human complexes, in vitro and in vivo, will provide important new insights into the mechanisms that control promoter-targeted chromatin modifications and that coordinate the transcription regulation of a selected group of genes during development, cell proliferation, and differentiation.