In Vivo Functional Analysis of the Histone 3-like TAF9 and a TAF9-related Factor, TAF9L*

The majority of the TATA-binding protein (TBP)-asso-ciated factors (TAFs) that constitute transcription factor II D (TFIID) contain histone fold motifs (HFMs). Our previous results utilizing DT40 cells containing a conditional TAF9 allele indicated that the histone 3-like TAF9 is essential for cell viability but largely dispensable for general transcription. In this study, we investigated further the role of TAF9 structural domains in TFIID integrity and cell growth and the functions of a TAF9-related factor, TAF9L. We first show that TAF9 depletion se-verely disrupts TFIID, indicating that the observed ongoing transcription is initiated with at least partially TAF-free TATA-binding protein. We also provide evidence for specific roles of TAF HFMs, highlighting the functional significance of HFM specificity observed in vitro and, importantly, of the TAF9-histone 3 similarity. Although we provide evidence that TAF9 and TAF9L are partly redundant, RNA interference experiments suggest that TAF9L is essential for HeLa cell growth. Strik-ingly, we provide evidence that TAF9L plays a role in transcriptional repression and/or silencing.

Initiation of transcription mediated by RNA polymerase II (RNAP II) 1 requires a number of general transcription factors (GTFs), among which TFIID is the major core promoter recognition factor. TFIID is composed of highly conserved subunits, including the TATA-binding protein (TBP) and about 12-14 TBP-associated factors (TAFs) (1)(2)(3)(4). Despite their diverse sequence and structural features, most TAFs (9 of 14 in yeast) contain domains known as histone fold motifs (HFMs) that display varying degrees of similarity with canonical histone core domains (5). TAF HFMs have been shown to mediate highly specific pairwise TAF-TAF interactions in cocrystal structures, in vitro interaction assays, and coexpression analyses (5)(6)(7)(8)(9). Furthermore, histone-like TAFs were shown in immunolabeling/electron microscopy analysis to colocalize to specific TFIID subdomains (10,11) in a pattern consistent with the above pairwise TAF interactions (12). Additionally, four yeast histone-like TAFs, TAF6/9 (H4/H3-like) and TAF4/12 (H2A/ H2B-like), were reconstituted into a histone octamer-like complex in vitro (13). However, the roles of the HFMs in TFIID structure and function in vivo are still unclear for most TAFs, especially in vertebrates. For example, the highly conserved arginine residues in histones that contact DNA are missing in TAFs (14), suggesting that histones and histone-like TAFs contact DNA differently.
A subset of the histone-like TAFs were also found to be integral components of histone acetyltransferase (HAT) protein complexes, including SAGA in yeast and GCN5, PCAF, TFTC, and STAGA in vertebrates (2,3,15,16). These HAT complexes are larger than TFIID, containing GCN5 or PCAF HAT activity and other factors likely involved in diverse nuclear processes (17,18). Importantly, however, histone-like TAFs that are missing in the vertebrate HAT complexes are replaced with a different set of HFM-containing subunits, and the specific pairwise patterns mediated by TAF HFMs are well conserved in these HAT complexes (5,13). Furthermore, electron microscopy images of TFTC showed an overall shape and subdomain organization similar to TFIID (11). These findings together strongly suggest that histone-like TAFs play important roles in maintaining the overall structures of TFIID and HAT complexes.
More recently, several TAFs (including TAF4, TAF6, and TAF9) and TBP were also found in yet another multisubunit complex, the Polycomb repressive complex 1 (PRC1) (19), in Drosophila. PRC1 functions to maintain transcriptional repression, or silencing, of certain regulatory genes, such as the HOX genes in flies and vertebrates (20,21). Chromatin immunoprecipitation analysis revealed the presence of GTFs on PRC1repressed promoters (22), suggesting that promoter-bound GTFs are locked in a transcriptionally inactive conformation and that TBP and TAFs found in PRC1 could serve as a link between target promoters and PRC1. However, the precise TAF composition of PRC1 and the role of TAFs in transcriptional repression await further experiments.
In addition to the core TAFs that exist in TFIID and related complexes, a number of variant TAFs have also been uncovered in metazoans. Some variant TAFs are derived from the same genes that encode core TAFs. For example, in response to apoptotic stimuli, a human TAF6 splicing isoform is induced and incorporated into a specialized TAF9-free TFIID required for an altered transcriptional program (23). In addition, a modified form of TAF9 was detected in both PCAF and STAGA but not in TFIID (18,24). Alternatively, distinct genes encode many variant TAFs. Unlike core TAFs that display ubiquitous expression patterns (25), these TAFs are typically highly expressed only in selective tissues (26,27). Disruption of one such tissue-specific TAF, TAF4B, in mice caused defective ovarian development (26). Furthermore, although yeast SAGA contains TAF5 and TAF6 (28), the human HAT complexes instead utilize related factors, TAF5L and TAF6L (17,18,24), and Dro-sophila TAF5L was shown to be required for normal spermatid differentiation (29). In addition, an apparent TAF9-related factor was found in two murine cell lines (30,31), and a systematic examination of the human genome uncovered six TAF9-related sequences (32).
Previously we constructed a conditional TAF9 knockout cell line in chicken DT40 cells (DT40-TAF9) (33). We provided evidence that TAF9 was not generally required for RNAP IImediated transcription in vivo, although depletion of TAF9 resulted in codepletion of many other TAFs and apoptotic cell death. Here we use DT40-TAF9 cells to investigate further the properties and functions of TAF9. We show that depletion of TAF9 causes disruption of TFIID, illustrating both the key role of TAF9 in TFIID structural integrity and the functional competence of TFIID with greatly reduced TAF content. Complementation analyses with various TAF9 chimeras reveal stringent sequence requirements of the HFM, highlighting the significance of HFM specificity and of the TAF9-H3 similarity. Human TAF9 and the related factor, TAF9L, both restored essential functions of cTAF9 in DT40 cells. To study TAF9L expression further, we show that reduction of hTAF9L expression in HeLa cells by small interfering RNAs (siRNAs) results in cell death. Interestingly, reduction of hTAF9L expression significantly induced transcription from several promoters, suggesting that TAF9L functions in transcriptional repression.

MATERIALS AND METHODS
Plamid Constructs-Mutant or chimeric cDNAs used in complementation experiments were generated by PCR. Human and Drosophila TAF9 cDNAs were kind gifts from Dr. R. Tjian. The HFM from cTAF9 and the CR from dTAF9 were subcloned into pKS(ϩ) (Stratagene) via three-fragment ligation to create cTAF9(HFM)-dCR. To make cDNAs encoding chimeric TAF9 proteins containing HFMs from dTAF9, hTAF12, hTAF6, or chicken H3, the flu tag sequence and the cTAF9 CT region were first inserted into pKS(ϩ) to make pflu-CT. HFMs, either PCR-amplified from cDNAs or cloned by RT-PCR, were inserted into pflu-CT. Single and triple point mutants of cTAF9 were generated using the QuikChange mutagenesis kit (Stratagene). The chimeric hTAF6 cDNA containing cTAF9 HFM (cTAF9-⌬hTAF6) was created by replacing the dTAF9 CR region in the cTAF9(HFM)-dCR with a fragment of hTAF6 cDNA (kindly provided by Drs. P. Vikas and R. Roeder) corresponding to amino acids 89 -677. Human TAF9L cDNA was a kind gift from Dr. H. Song. The above cDNAs were subcloned into expression vectors pAPSV-zeo (60) and/or PA-puro (33), except that cTAF9-⌬hTAF6 was inserted into the expression vector pSV-Brs. The luciferase reporter constructs containing core or enhancer-containing promoters and the expression vector encoding TAF9 fused to three copies of the FLAG tag at the C terminus are described elsewhere (60).
Cell Culture and Transfections-Chicken DT40-TAF9 cells were maintained and transfected by either electroporation as previously described (33) or LipofectAMINE (Invitrogen) according to the manufacturer's manual. HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Transient transfections of HeLa and 293 cells were carried out using standard CaPO 4 method. For stable transfections of HeLa cells, 10 g of linearized plasmid DNAs were incubated with cells from one 100-mm dish at 70% confluency in 0.7 ml of Hepes-buffered saline, and mixtures were electroporated at 250 V/950 uF. Two days later, cells were plated out to eight 100-mm dishes in selective media. Blasticidin (Calbiochem) and Zeocin (Invitrogen) were used at 10 g/ml and 200 g/ml, respectively.
Protein and RNA Analyses-Western blotting and RNase protection analyses were carried out as previously described (33). Northern blotting analysis was carried out by standard procedures. Briefly, 15-50 g of total RNAs were run on formaldehyde-agarose gels, transferred to nylon membranes, and hybridized with 2 ϫ 10 6 cpm/ml probe. Luciferase reporter assays were carried out as recommended (Promega) using a luminometer (Berthold). Briefly, 24 -36 colonies of transfected HeLa cells were picked by cloning disks (Scienceware) from each transfection and analyzed for luciferase expression. Selected clones, with either median or high luciferase activities, were treated with siRNAs for 3.5 days prior to luciferase assays.
Anti-TBP antibodies, including mAb 3G3 (0.5 l of ascites/IP) (35) and the polyclonal sc-273 (4 g/IP, Santa Cruz Biotechnology), were cross-linked to protein G or A-Sepharose beads, respectively, using dimethylpimelimidate as described (36). The above mini-nuclear extracts or WCEs were diluted 1:1 with an IP binding buffer (phosphatebuffered saline with 0.04% bovine serum albumin, 0.1% Nonidet P-40, and protease inhibitors) and incubated with antibody-conjugated beads for 4 h to overnight. After washing four times with the IP binding buffer, the beads were resuspended in sample buffer and loaded onto SDS-PAGE gels.
Efficient expression of proteins of interest was confirmed in all experiments by Western blots, and in all cases at least two independent clones expressing the same exogenous protein at levels similar to or higher than those of TAF9 were subjected to co-IP analysis. TBP IP Western blots were routinely carried out, and similar IP efficiency was confirmed in all experiments.
SiRNA Transfections and Analyses of Cell Growth-Design, reconstitution, and transfection of siRNAs were conducted essentially as described (37) with several modifications. Briefly, a TAF9L-specific nucleotide sequence just downstream of the conserved region was chosen for designing the siRNA specific for TAF9L (Dharmacon). After reconstitution, 4 l of 20 mM siRNA duplex were transfected with 4 l of Oligofectamine (Invitrogen) into HeLa cells that were seeded at 2 ϫ 10 4 /well in 24-well plates 24 h before. Cotransfection of plasmid DNAs and siRNAs was carried out essentially as described (38). Microscopy images were taken using a NIKON Diaphot 300 microscope. DNA fragmentation analysis was performed as previously described (39).

TAF9 Depletion Disrupts TFIID Complexes-Previously we
showed that depletion of TAF9 in DT40-TAF9 cells resulted in significant codepletion of several other TAFs (33). An important question was whether this also reflected the composition of TFIID: is TFIID disrupted by TAF9 depletion, or does it stay relatively intact such that only the levels of free TAFs are reduced? To distinguish between these possibilities, coimmunoprecipitation (co-IP) experiments using an anti-TBP mAb (35) were performed with extracts from DT40-TAF9 cells grown in the presence of tet (1 g/ml) for 0, 36, and 72 h, and precipitates were analyzed by Western blotting (Fig. 1). As observed previously (33), only a slight reduction in TBP protein levels was observed upon TAF9 depletion (Input, top panel), whereas TAF7 (and several other TAFs) protein levels were significantly reduced. Among the TAFs tested, only TAF5 protein levels were not affected, even after 72 h of incubation in 10 FIG. 1. Depletion of TAF9 disrupts TFIID. DT40-TAF9 cells were incubated in 1 g/ml tet for the time indicated. Nuclear extracts were immunoprecipitated with the anti-TBP mAb 3G3, and aliquots of input and precipitates (IP) were analyzed by Western blot analysis using 3G3, anti-flu antibody (Covance), and TAF5-and TAF7-specific antibodies. Molecular mass markers and identities of proteins detected are indicated. g/ml tet ( Fig. 1 and data not shown) (33). Importantly, depletion of TAF9 from cell extracts coincided with its efficient removal from TFIID (IP). Strikingly, not only TAF7 but also TAF5 showed significant dissociation from TBP after 36 h in tet and were nearly undetectable by 72 h. This co-dissociation of TAF7 was especially remarkable given that it does not seem to interact directly with TAF9 (40). These results suggest that depletion of TAF9 significantly compromises TFIID complex integrity, likely leading to dissociation of the majority of TAFs from TBP. Coupled with our previous results indicating that RNAP II transcription is not significantly reduced after 72 h of tet treatment, these findings indicate that an intact TFIID complex is not generally required to maintain high levels of transcription.
Dissection of TAF9 Functional Domains-The above results underscore the important role of TAF9 in TFIID complex integrity, and despite the fact that it is not generally required for transcription, TAF9 is essential for cell viability (33). To understand further the in vivo functions of TAF9, we next examined the role of TAF9 structural domains in supporting cell viability and TFIID complex assembly. To this end, we employed DT40-TAF9 cells for the following genetic complementation analysis: cDNAs encoding proteins of interest were cloned into an expression vector with a selectable marker (zeocin, or zeo), and DT40-TAF9 cells were transfected with these expression vectors and selected either in tet for surviving colonies or in zeo for clones expressing the protein of interest, which were then grown in tet to determine cell viability. Both selection methods gave consistent results in all cases. As shown in Fig. 2, mock (no DNA) or empty vector transfections resulted in no surviving colonies in tet, whereas chicken TAF9 (cTAF9) itself gave rise to high numbers of surviving colonies, validating this approach. To examine TFIID incorporation, co-IP experiments with the anti-TBP mAb were carried out with extracts prepared from clones expressing the protein of interest at levels comparable with or higher than those of TAF9. High and very similar IP efficiency in all experiments was confirmed by TBP Western blot analysis ( Fig. 2A and data not shown).
The C-terminal tail region of vertebrate TAF9 (TL), although accounting for almost one half of the protein and highly conserved between chicken and human, is not similar in sequence to the corresponding region in Drosophila TAF9 and is completely missing in yeast TAF9. When a C terminus-truncated cTAF9 mutant (cTAF9M1) was expressed in DT40-TAF9 cells, the cells grew normally in tet, and cTAF9M1 was incorporated efficiently into TFIID ( Fig. 2A). In sharp contrast, deletion of the highly conserved HFM flanking region (CR), despite its relatively small size and lack of detectable secondary structure, resulted in a protein that was completely unable to rescue cell viability and was not incorporated into TFIID (cTAF9M2). A more limited truncation of the CR and complete removal of the C-terminal tail (cTAF9M3) resulted in a slow growth phenotype and significantly decreased TFIID incorporation efficiency relative to that of cTAF9. These results indicate that the CR, but not the C-terminal tail, plays an essential role in cell growth and TFIID complex assembly.
Next we examined evolutionary conservation of TAF9. As expected, the highly similar human TAF9 (hTAF9; 86% identical in amino acid sequence) protein supported cell growth in tet and TFIID incorporation. Surprisingly, however, the number of hTAF9 colonies selected in tet was consistently lower than observed with cTAF9 despite similar protein expression levels (compare Fig. 2, A and B; data not shown), indicating that the minor sequence differences between hTAF9 and cTAF9 compromise hTAF9 function in DT40-TAF9 cells. Drosophila TAF9 (dTAF9), with significantly reduced sequence similarity (41% identity), failed to support cell growth and was not incorporated into TFIID (Fig. 2B). dTAF9 was shown to interact with hTAF6 in vitro (41); our results thus suggest that other interactions are required for incorporation into TFIID. The inability of dTAF9 to function was not because of any negative effects from the long, non-conserved C-terminal tail, because a dTAF9 mutant lacking the tail also did not support cell growth or TFIID incorporation (Fig. 2B). Interestingly though, a chimeric dTAF9 containing the CR and C-terminal tail from cTAF9 (dTAF9-cCT) supported cell survival, as did a chimeric cTAF9 containing the CR from dTAF9 (cTAF9-dCR), although both chimeras displayed reduced TFIID incorporation efficiency and yielded fewer colonies than did cTAF9 (Fig. 2C, compare IP lanes with input lanes). These findings suggest the existence of a functional cooperation between the HFM and CR.
Previous coexpression experiments in E. coli have established that specific pairwise TAF-TAF interactions are mediated through the HFMs (5). To understand further how specific HFM interactions contribute to TFIID assembly, we next examined whether substituting the TAF9 HFM with HFMs from other TAFs generates functional TAFs. HFMs from hTAF12 (H2B-like) and hTAF6 (H4-like) were used to replace the cTAF9 HFM to make hTAF12-CT and hTAF6-CT, respectively (Fig. 3A). Although cells expressing these chimeras displayed normal growth rate in the absence of tet, both chimeric proteins failed to support cell growth upon cTAF9 depletion, emphasizing the essential role of the cTAF9 HFM in cell viability. Unexpectedly, however, both chimeras associated with TBP under normal growth conditions, indicating that they can be incorporated into TFIID. Significantly, cTAF9 depletion dramatically reduced the levels and TBP association of both chimeras (Fig.  3A), similar to the behavior of most other natural TAFs ( Fig. 1; Ref. 33), indicating that incorporation of the chimeras into TFIID was TAF9-dependent. By contrast, a chimeric hTAF6 with the cTAF9 HFM replacing the TAF6 HFM (cTAF9-⌬hTAF6) failed to rescue viability or, significantly, to associate with TBP (Fig. 3B). Furthermore, cotransfection of cTAF9-⌬hTAF6 and hTAF6-CT did not result in surviving colonies in tet or TFIID incorporation (data not shown). These results together indicate that in some cases distinct HFMs can function generically with respect to directing incorporation into TFIID but that they have distinct, TAF-specific functions beyond this.
TAF9 and H3 HFMs are Functionally Related-Although the cTAF9 HFM is most similar to that of histone 3, the HFMs from these two proteins display only low sequence identity (28%), and, for example, the conserved Arg residues in histones that contact DNA are absent in all TAFs (14). To characterize the functional difference of these HFMs, we first replaced the HFM in cTAF9 with chicken H3 HFM (H3-CT). Remarkably, H3-CT associated with TBP under normal growth conditions (Fig. 3A). Additionally, unlike other chimeric and natural TAFs, a significant amount of H3-CT (roughly 1/3 of the level in untreated cells) was detected and shown to remain associated with TBP in tet-treated cells, suggesting a limited ability of H3-CT to be incorporated into TFIID independent of TAF9. Consistent with this, DT40-TAF9 cells expressing H3-CT, although not viable in the presence of tet, died more slowly after day 4 (by ϳ16 -24 h) than did untransfected DT40-TAF9 cells (Fig. 3D) or cells expressing chimeric TAF9 proteins containing HFMs from TAF6 or TAF12 (data not shown). These results provide strong support for the significance of the similarity between specific TAFs and histones.
The inability of H3-CT to restore fully cell viability and TFIID integrity in the absence of cTAF9 could have resulted either from low overall sequence similarity between their HFMs or from the absence of specific conserved residues at potentially critical positions. To address this, we next determined the effects of specific substitutions in the TAF9 HFM that increase the similarity with histones. Besides the two histone-specific Arg residues that contact DNA, we also analyzed an Ala (TAF9) to Thr (H3) difference in ␣2 because this residue is almost completely conserved among TAFs but not in histones (5,6,14). Two TAF9 mutants, one containing a single point mutation (cTAF9-A54T) and the other a triple mutant with the addition of A10R and E30R double mutations (cTAF9triple) were constructed and analyzed. Strikingly, both mutants behaved indistinguishably from cTAF9 in cell viability assays and TBP association when cTAF9 was depleted (Fig.  3C). Results from chromatin immunoprecipitation analysis also showed no differences between wild-type and mutant proteins in promoter occupancy (data not shown). Taken together, these results indicate that these conserved differences between TAF9 and histone HFMs are not important for the functions tested here.
TAF9 Protein Levels Are Tightly Regulated-Previously we observed that flu-cTAF9 levels in DT40-TAF9 cells were only 1.5-fold greater than the levels of endogenous TAF9 despite highly overexpressed (ϳ50-fold) mRNA levels (33), suggesting that a post-transcriptional mechanism tightly regulates TAF9 protein levels. This would predict that in cells expressing both cTAF9 and a functionally related protein, the combined levels of the two proteins are also regulated. As shown in Fig. 4A (top  panel), cTAF9M1 levels were relatively low when cells also FIG. 4. TAF9 protein levels are tightly regulated. A, post-transcriptional control of TAF9 expression. Top panels, Western blot analysis of TAF9 and two related proteins. DT40-TAF9 cells expressing cTAF9M1 (left) or dTAF9 (right) were grown in the absence (Ϫ) or presence (ϩ) of tet for 72 h and subjected to anti-flu Western analysis. Bottom panel, cTAF9M1-expressing cells were grown as above and subjected to RNase protection analysis; probe and protected products are indicated. B, tet titration analysis. Top panel, DT40-TAF9 cells were grown in tet at the indicated concentrations for different times and whole-cell extracts (WCE) were prepared and analyzed by Western blot using anti-flu and anti-TAF7 antibodies. Bottom panel, DT40-TAF9 cells were grown in the absence or presence of 5 ng/ml tet for 3 days, and anti-flu Western analysis was carried out using the indicated amounts of WCEs from these cells. C, growth curves of DT40-TAF9 cells in the presence of different concentrations of tet. Cells were split every 2 days at the starting density of 2 ϫ 10 5 cells/ml. expressed flu-cTAF9, but when cTAF9 was depleted cTAF9M1 protein accumulation was enhanced to a level comparable with the combined protein levels in the absence of tet. RNase protection analysis showed constant cTAF9M1 transcript levels in the presence or absence of tet, consistent with a post-transcriptional control. In sharp contrast, levels of dTAF9, which neither assembled into TFIID nor supported cell growth in tet, were significantly elevated relative to cTAF9 and minimally affected by TAF9 depletion. These results indicate that a tight mechanism limits accumulation of functional TAF9, perhaps with TFIID incorporation as at least part of the signal.
The above results also suggest a close relationship between TAF9 protein levels and cell growth. To examine this more directly, we carried out tet titration experiments to characterize further the link between TAF9 depletion, codepletion of other TAFs, and cell growth. At 1 g/ml tet, TAF9 and to a lesser degree TAF7 were efficiently depleted, and cells started to die after 60 -72 h (Fig. 4B) (33). Incubation at 2 ng/ml tet had no effect on either TAF protein levels or cell growth (not shown), whereas 8 ng/ml tet effected a rapid, although slightly slower than with 1 g/ml, reduction of TAF9/TAF7 protein levels, and eventually led to cell death (Fig. 4, B and C). At 5 ng/ml tet, however, cells continued slow growth and exhibited moderate cell death ( Fig. 4C and data not shown). Significantly, these growth phenotypes coincided with a partial reduction of TAF9, and TAF7, protein levels (Fig. 4B, top). Western analysis showed that TAF9 levels were roughly 20% of time 0 (Fig. 4B, lower panel) or 30% of endogenous TAF9 levels in wild-type DT40 cells (33). These results indicate that TAF protein levels are tightly modulated and correlate directly with cell growth.
Human TAF9L Is Essential for Cell Growth-Systematic examination of the human genome uncovered six hTAF9-related sequences (32). Whereas five of these sequences are likely pseudogenes, because they lack introns and/or contain frameshift mutations (data not shown), two nearly identical cDNA sequences of the remaining gene, termed TAF9L, were found in the GenBank TM data base; two murine homologues of hTAF9L have also been identified. The sequences of TAF9 and TAF9L are highly similar throughout their entire coding regions (78% identity; Fig. 5A). Although alterations of TAF9L expression patterns have been correlated with apoptosis or differentiation in two murine cell lines (30,31), functions of TAF9L are not FIG. 5. Human TAF9L can functionally replace cTAF9 in DT40-TAF9 cells. A, protein sequence alignment. Amino acid sequences of human (h) and rat (r) TAF9L and of human and chicken (c) TAF9 were aligned using Clustal W software. At those positions where at least three proteins have identical residues, amino acids are marked in dark gray. At positions where residues are only identical in TAF9 or TAF9L sequences, amino acids are marked in light gray. The overlined sequence denotes the region corresponding to the siRNA oligos in Fig. 6A. B, analysis of DT40-TAF9 cells expressing hTAF9L. Left, anti-TBP co-IP analysis. Cells were grown in tet for 72 h. Co-IP with no antibody (mock) or mAb 3G3 was performed, and precipitates were analyzed by anti-flu Western analysis. Right, two DT40-TAF9 clones expressing hTAF9L were grown in the absence (Ϫ) or presence (ϩ) of tet for 72 h and analyzed by anti-flu Western blotting using WCEs (top) or by RNase protection assay (bottom). known. Low stringency Northern or RT-PCR analysis did not detect expression of a putative cTAF9L in DT40-TAF9 cells (data not shown).
To begin analysis of TAF9L functions, we first asked whether exogenous expression of hTAF9L rescues DT40-TAF9 viability in the presence of tet. DT40-TAF9 cells expressing hTAF9L were obtained as above. In the presence of tet, TAF9L was efficiently incorporated into TFIID (Fig. 5B, left), and the cells grew normally (not shown). Additionally, hTAF9L mRNA expression in these cells remained constant in the absence or presence of tet, whereas hTAF9L protein levels in tet-treated cells appeared to increase (Fig. 5B, right). These results indicate that exogenous hTAF9L expression can fully restore cTAF9 functions in DT40-TAF9 cells, indicating that TAF9L can function as a bona fide TAF.
We next asked whether hTAF9L is essential for growth of human cells. To this end, we designed siRNA oligos (37) corresponding to an hTAF9L but not hTAF9 mRNA sequence (Figs. 5A and 6A, denoted regions). HeLa cells were transfected with the TAF9L siRNA duplex (see "Materials and Methods"), and total RNAs were purified for Northern analysis. As shown in Fig. 6B, TAF9L siRNA transfection significantly reduced TAF9L mRNA levels (top band) but had little effect on TAF9 mRNA (bottom band; from cross-hybridization with the TAF9L probe). By contrast, a control siRNA specific for luciferase (luc) showed no effect on either TAF9 or TAF9L mRNA accumulation. Visually, cells transfected with TAF9L siRNA showed significant levels of cell death (Fig. 6C); concomitantly, DNA fragmentation assays showed increasing amounts of fragmented DNA in cells treated with TAF9L siRNA for 3 and 4 days (Fig. 6D, top panel, lanes 4 and 5), whereas control (untransfected) and cells transfected with luc siRNA showed no effects on cell growth (Fig. 6C) and DNA fragmentation (Fig.  6D, lanes 1 and 2). These results suggest an essential role of hTAF9L in human cell growth.
Next we examined the effect of reduced TAF9L levels on TAF-containing complexes (Fig. 6E). In agreement with minimal effects of TAF9L siRNA on TAF9 mRNA levels, TAF9 protein levels remained constant. Moreover, TAF9L siRNAtreated cells maintained normal levels of TBP and TAF7, strongly suggesting that TFIID complex integrity was not affected. In contrast, two subunits of GCN5 HAT complexes, Gcn5-S and TAF12 (also a subunit of TFIID), displayed significantly reduced protein levels after 3 and 4 days, respectively, suggesting a possible role of TAF9L in GCN5 complex assembly.
Previously we constructed a HeLa cell line stably expressing hTAF9 with a C-terminal 3ϫ FLAG epitope tag under the control of the cytomegalovirus promoter (FLAG-TAF9 cells) (60). To investigate whether elevated levels of TAF9 can rescue cell death induced by TAF9L knockdown, we treated FLAG-TAF9 cells with TAF9L siRNA. Strikingly, treatment with the TAF9L siRNA, but not the luc siRNA, strongly induced accumulation of FLAG-TAF9 protein but had no effects on endogenous TAF9 (Fig. 6E and data not shown). FLAG-TAF9 mRNA levels were similarly induced in TAF9L siRNA-treated cells (Fig. 6D, bottom panel). Furthermore, this induction of FLAG-TAF9 correlated with delayed cell death: although FLAG-TAF9 cells showed increasing levels of fragmented DNAs after 3 and 4 days of TAF9L siRNA treatment (Fig. 6D, top panel, lanes 6  and 7), the levels of fragmented DNAs were consistently lower than in HeLa cells subject to the same treatment (compare lanes 4 and 6, and lanes 5 and 7). These results illustrate an essential function of TAF9L in cell growth that can only be partially compensated by overexpression of TAF9.
TAF9L Functions in Transcriptional Repression-Next we further examined the basis for the surprising transcriptional induction of FLAG-TAF9 transcription by TAF9L knockdown. As noted above, expression of several endogenous genes, including TAF9, was not affected or was reduced by TAF9L siRNA treatment. Additional Northern and Western analyses revealed that TAF9L knockdown had no detectable effect on expression of endogenous poly(A) polymerase, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and actin genes (Fig. 6,  B and E), and Northern analysis using a poly(A) probe showed constant levels of total mRNA (data not shown). These results indicate that TAF9L does not generally repress accumulation of transcripts from endogenous genes. We next asked whether TAF9L could repress transcription from other stably integrated promoters. Previously we established HeLa cell lines stably transfected with luciferase reporter constructs containing various core or enhancer-containing promoters (60) (see "Materials and Methods"). Cells expressing luciferase from specific promoters at typical levels, as measured by median expression of individual clones, were treated with control and TAF9L siRNAs for 3.5 days. In this case, the luc siRNA acted as a positive control and reduced luciferase activity to ϳ5% in all cases tested (Fig. 7A, luc lanes), whereas another control siRNA, specific for a SR protein (SRp38), had no effect on luciferase expression. Strikingly, although the TAF9L siRNA showed essentially no effect on luciferase activities from the adenovirus major late (MLP) and Drosophila TATA-less jockey core promoters (Fig. 7A, top panels), luciferase expression from corresponding enhancer-containing promoters was significantly induced (Fig. 7A, bottom, SV-MLP and IRF-1), ranging from ϳ7-fold for IRF-1 (core promoter region similar to jockey) to ϳ14-fold for SV-MLP (which contains the SV40 72-bp enhancer repeats). Interestingly, when cells that expressed luciferase at levels higher than the median (ϳ4 -5-fold) were treated with TAF9L siRNA, induction was less pronounced, around 3-fold for both SV-MLP and IRF-1 (Fig. 7A, bottom, SV-MLP(high) and IRF-1(high)). RNase protection analysis confirmed induction of luciferase mRNA levels in TAF9L siRNA-treated cells (Fig. 7B, top). These results indicate that TAF9L is involved in repression of enhancer-dependent activated transcription from several genes.
Unlike the constitutively expressed endogenous genes examined in Fig. 6, B and E, the endogenous IRF-1 gene is inducible and displays low transcriptional activity before induction by interferon (42). We therefore hypothesized that TAF9L is required to maintain the transcriptionally inactive state of IRF-1 prior to induction. To address this, and to rule out the possibility that TAF9L-dependent repression only acts on exogenous genes, we analyzed endogenous IRF-1 mRNA levels. Strikingly, a strong induction of endogenous IRF-1 transcription was observed in cells treated with TAF9L, but not SRp38, siRNA (Fig.  7B, bottom). Together these results indicate that, although partly redundant with TAF9, TAF9L is required for transcriptional repression of certain inducible and enhancer-driven genes.

DISCUSSION
Our previous results showed that depletion of TAF9 caused significant codepletion of several other TAFs, but general RNAP II transcription was not markedly affected (33). Here we showed that TAF9 depletion, in fact, causes severe disruption of TFIID, indicating that for many genes ongoing transcription does not require intact TFIID. Chromatin immunoprecipitation analyses have also revealed complete removal of TAF9 from various promoters after 72 h in tet, and promoter occupancy levels of TAF6 and TAF7 were also found to be significantly reduced on several promoters (60). Davidson and co-workers FIG. 6. TAF9L is essential for HeLa cell growth. A, sequence of the siRNA specific for hTAF9L. The underlined sequence corresponds to hTAF9L siRNA. Sequence of the corresponding region in hTAF9 is included for comparison. B, Northern blot analysis. HeLa cells were treated with mock, luciferase (luc), and TAF9L (9L) siRNAs for the indicated times. Total RNAs were prepared for Northern analysis using hTAF9L (top panel; the probe weakly hybridized with hTAF9), poly(A) polymerase (PAP), and glyceraldehyde-3-phosphate dehydrogenase probes. C, images of HeLa cells siRNA-treated for the indicated times. D, top panel, DNA fragmentation analysis. HeLa or FLAG-TAF9 cells were treated with the indicated siRNAs for 3 or 4 days. Genomic DNAs were purified and run on a 1.2% agarose gel. Bottom panel, FLAG-TAF9 cells were treated with the indicated siRNAs for 3 days. Total RNAs were prepared and subjected to RNase protection analysis using a probe specific for FLAG-TAF9 mRNA. E, Western analysis. HeLa and FLAG-TAF9 cells were treated as in panel B. WCEs were prepared and used for Western analysis using antibodies specific for the indicated proteins. The apparent molecular masses for Gcn5-S, TAF12, and Actin are 47, 20, and 42 kDa, respectively. (43) recently showed that TBP-disrupted murine cells also sustain high levels of RNAP II transcription. These results together reveal unexpected plasticity, and perhaps redundancy, of TFIID functions in RNAP II transcription. Despite this, levels of TAF9 (and TBP as well; Ref. 44) are very tightly regulated. As shown here, reduction of TAF9 protein levels results in slow cell growth, and depletion causes apoptotic cell death (33). In this regard, it is notable that a specialized, TAF9-free TFIID (TFIID), containing a variant TAF6 (TAF6␦) that cannot interact with TAF9, is induced by apoptotic stimuli and may direct expression of specific apoptotic genes (23), perhaps providing part of the explanation why TAF9 levels must be so tightly regulated. Overexpression of TAF6, on the other hand, was shown to have no apparent effects on cell growth and development in Drosophila (45). Taken together, these results indicate the pivotal roles played by TAF9 in both TFIID integrity and cell growth. Below we discuss how our studies on TAF9 structural domains and the unexpected properties of TAF9L contribute to understanding TAF9 function.
Our results provide support for the physiological significance of a TAF HFM interaction "code" that has emerged from in vitro studies (5). For example, our finding that a chimeric TAF9 containing the H3 HFM partially compensates for TAF9 deple-tion establishes a functional similarity between the H3 and TAF9 HFMs, extending previous observations that the H3 HFM interacts in vitro with the HFM from the TAF9 partner, TAF6 (6,7,14). Our results also indicate that the TAF9 HFM is specifically required for TFIID integrity and cell growth, because chimeric TAF9s containing HFMs from TAF6 and TAF12 did not rescue TAF9 functions. Interestingly, however, rather than replacing TAF9 in TFIID, these TAF9 chimeras behaved like other natural TAFs, displaying TAF9-dependent accumulation and TFIID incorporation. It is plausible that the chimera containing the TAF12 HFM replaces TAF12 in TFIID, because the TAF12 HFM encompasses the entire highly conserved region and is sufficient to support yeast cell growth (46 -48). Similarly, the chimera containing the TAF6 HFM might also enter TFIID via TAF6 HFM interactions with HFMs from TAF9 and TAF12, although the absence of the TAF6 non-HFM region likely prevents interactions between the natural TAF6 and other TFIID subunits (49). Our data therefore suggest either that these additional interactions are not absolutely required for TFIID integrity, or, in light of the recent finding that TFIID contains at least two molecules of each histone-like TAF (12), that a natural TAF6 might cooperate with the chimeric protein in the same TFIID complex. Together, our results illustrate the ability of TAF HFMs to func- FIG. 7. Transcriptional induction in TAF9L siRNA-treated cells. A, HeLa cells were stably transfected with luciferase reporter constructs containing the indicated core or enhancer-containing promoters. Individual clones were isolated and assayed for luciferase activity. Cells with median or high luciferase activities as indicated were treated with the indicated siRNAs for 3.5 days and luciferase activities were determined. Error bars represent S.D. from two to three independent experiments. B, RNase protection. HeLa cells expressing luciferase from the IRF-1 promoter or normal HeLa cells were treated with mock, SRp38, or TAF9L siRNAs for 3.5 days. Total RNAs were purified and subjected to RNase protection analysis using probes specific for luciferase (top) or endogenous IRF-1 mRNA (bottom). tion to a certain degree in a heterologous context and, more importantly, provide functional significance for the specificity of TAF HFM-HFM interactions.
A number of TAF-related factors have been identified in metazoans (32,50). Although TAF5L and TAF6L are conserved from Drosophila to human and share modest sequence similarity with their respective core TAFs (46 and 24% identity, respectively; Ref. 24), TAF9L is not present in Drosophila but is highly similar to TAF9 (78% identity; Refs. 30 -32), suggesting that a relatively recent gene duplication event gave rise to TAF9L. Given the high sequence similarity, it was not surprising that our results showed partly redundant functions of TAF9 and TAF9L. Unexpectedly, though, both TAF9 and TAF9L are essential for cell growth, and knockdown of TAF9L did not elicit a compensatory increase in TAF9 expression, suggesting functional divergence. This is in contrast to the case where gene targeting of PCAF in mice had no apparent deleterious effects on development, likely reflecting a significant induction of the related Gcn5/PCAF-B (51). Consistent with distinct functions, knockdown of TAF9L had no effects on levels of TFIID-specific subunits, suggesting that TAF9L is not an integral component of TFIID. Another possibility is that TAF9L is incorporated into the TAF9-free TFIID that functions during apoptosis (23), especially in light of the previously observed TAF9L induction during apoptosis of rat PC12 cells (30,52). However, TAF9L, like TAF9, did not interact with the truncated HFM of TAF6␦, 2 suggesting that TAF9L is unlikely to be directly involved in TAF6␦-dependent function.
TAFs are generally regarded as positive cofactors in transcriptional regulation (3,16). Although certain TAFs were shown to function in derepression of specific genes in yeast (53), there has previously been no evidence that TAFs function in transcriptional repression. It was thus surprising that knockdown of TAF9L induced transcription from several stably integrated promoters and from an endogenous inducible promoter. Despite apparent sequence and functional heterogeneity in these promoters, transcription from them was apparently repressed to different degrees in normally growing cells prior to TAF9L knockdown. For example, transcription activity from the SV-MLP, containing the strong SV40 enhancer, was actually lower than that from the uninduced IRF-1 promoter and only ϳ20-fold greater than from the MLP core promoter. Interestingly, SV-MLP activity is ϳ20-fold higher in stably transfected DT40 cells, which as described above appear to lack TAF9L (60). Furthermore, our finding that TAF9L siRNA treatment induced expression to significantly lesser degrees in HeLa cell clones with higher luciferase activities, likely reflecting integration of reporter constructs into more transcriptionally active positions in the genome, suggests a more limited involvement of TAF9L in more actively expressed genes.
Our results suggest a possible role for TAF9L in both position effect-dependent silencing of exogenous genes and maintenance of transcriptional repression of uninduced genes. Transcriptional repression, or silencing, of stably integrated genes has been shown to result from heterochromatin packaging characteristic of the integration position in the genome (54 -57). In this regard, it is interesting that the Drosophila TAF9containing PRC1 complex has been implicated in maintenance of heterochromatin silencing such as in position effect variegation (19,21). Furthermore, GTFs and PRC1 subunits were found both to interact in vitro and to colocalize on target promoters (22). Although the analogous human complex did not seem to associate tightly with subunits of TFIID (58), loose and/or promoter-dependent TAF-PRC1 associations may occur and actually help PRC1 recruitment to and/or retention on core promoters (59). Given that TAF9L does not exist in Drosophila, our results are consistent with the hypothesis that TAF9L has replaced TAF9 in PRC1-dependent transcriptional silencing in vertebrates. However, alternative possibilities cannot currently be excluded. For example, in light of the observed reduction in Gcn5 protein levels by TAF9L knockdown, TAF9L might function as a subunit of a Gcn5-containing complex. In any event, it is remarkable that TAF9 and TAF9L are partly redundant despite performing what appear to be opposite functions.