Transforming Growth Factor-β Regulates DNA Binding Activity of Transcription Factor Fli1 by p300/CREB-binding Protein-associated Factor-dependent Acetylation*

Fli1, a member of Ets transcriptional factors, has been shown to be a negative regulator of collagen gene expression in dermal fibroblasts. Although Fli1 down-regulation is implicated in pathological matrix remodeling such as cutaneous fibrosis in scleroderma, very little is known about the post-translational mechanisms regulating Fli1 function. The aim of this study was to investigate the role of acetylation, one of the main post-translational regulatory mechanisms, in regulating Fli1 activity. We initially demonstrated that Fli1 is acetylated by transforming growth factor (TGF)-β1 in dermal fibroblasts. An in vivo acetylation assay using 293T cells revealed that Fli1 is mainly acetylated by the histone acetyltransferase activity of p300/CBP-associated factor (PCAF) at lysine 380. Acetylation of Fli1 resulted in a decreased stability of Fli1 protein. More importantly, reduced binding of acetylated Fli1 to the human α2(I) collagen (COL1A2) promoter was observed in DNA affinity precipitation and chromatin immunoprecipitation. Conversely, a Fli1 K380R mutant that is resistant to acetylation by PCAF showed increased DNA binding ability. Furthermore, PCAF overexpression reversed the inhibitory effect of Fli1 on TGF-β1-mediated COL1A2 promoter activity. In contrast, the Fli1 K380R mutant had a greater inhibitory effect on TGF-β1-induced COL1A2 promoter activity than wild-type Fli1, and PCAF failed to reverse this effect. These results indicate that PCAF-dependent acetylation of lysine 380 abrogates repressor function of Fli1 with respect to collagen gene expression. Furthermore, these data strongly suggest that the TGF-β-dependent acetylation of Fli1 may represent the principal mechanism responsible for the TGF-β-induced dissociation of Fli1 from the collagen promoter.

Fli1, a member of Ets transcriptional factors, has been shown to be a negative regulator of collagen gene expression in dermal fibroblasts. Although Fli1 down-regulation is implicated in pathological matrix remodeling such as cutaneous fibrosis in scleroderma, very little is known about the post-translational mechanisms regulating Fli1 function. The aim of this study was to investigate the role of acetylation, one of the main post-translational regulatory mechanisms, in regulating Fli1 activity. We initially demonstrated that Fli1 is acetylated by transforming growth factor (TGF)-␤1 in dermal fibroblasts. An in vivo acetylation assay using 293T cells revealed that Fli1 is mainly acetylated by the histone acetyltransferase activity of p300/CBP-associated factor (PCAF) at lysine 380. Acetylation of Fli1 resulted in a decreased stability of Fli1 protein. More importantly, reduced binding of acetylated Fli1 to the human ␣2(I) collagen (COL1A2) promoter was observed in DNA affinity precipitation and chromatin immunoprecipitation. Conversely, a Fli1 K380R mutant that is resistant to acetylation by PCAF showed increased DNA binding ability. Furthermore, PCAF overexpression reversed the inhibitory effect of Fli1 on TGF-␤1-mediated COL1A2 promoter activity. In contrast, the Fli1 K380R mutant had a greater inhibitory effect on TGF-␤1-induced COL1A2 promoter activity than wild-type Fli1, and PCAF failed to reverse this effect. These results indicate that PCAF-dependent acetylation of lysine 380 abrogates repressor function of Fli1 with respect to collagen gene expression. Furthermore, these data strongly suggest that the TGF-␤-dependent acetylation of Fli1 may represent the principal mechanism responsible for the TGF-␤-induced dissociation of Fli1 from the collagen promoter.
Fli1 (Friend leukemia integration-1) 3 is a member of the Ets (E26 transformation-specific) transcription factor family char-acterized by a winged helix-turn-helix DNA binding domain (ETS domain) that is responsible for nuclear targeting and specific binding to a DNA element containing a purine-rich GGA(A/T) core sequence (Ets binding site (EBS)) (1,2). Ets transcription factors are classified into subfamilies by the presence of several conserved domains for DNA binding and its location within the protein as well as additional sequence similarities (1,2). Fli1 and the closely related Erg belong to the ERG subfamily. Based on comparative gene expression profile analysis of various human cell lines, Fli1 is expressed at high levels in both endothelial and hematopoietic cells (3). Extensive in vitro studies as well as the data obtained from various Fli1 null mice support a crucial role of Fli1 in megakaryocytic differentiation and myelomonocytic, erythroid, and natural killer cell development (4 -7). Mice with a null mutation at the Fli1 locus die at day 11.5 of embryogenesis, due to a loss of vascular integrity, suggesting that Fli1 is involved in the regulation of genes critical for vascular remodeling (4). However, its target genes and its role in the vasculature have not been fully characterized.
Although Fli1 is present in a relatively limited amount in dermal fibroblasts, recent studies have shown that Fli1 plays a pivotal role in the regulation of extracellular matrix genes, including type I collagen (8 -10) and tenascin-C (11,12), extracellular matrix-degrading enzyme MMP-1 (13), and the multifunctional matricellular factor CCN2 (14). Most importantly, Fli1 has been shown to be a potent inhibitor of collagen biosynthesis in dermal fibroblasts, and its aberrant expression has been implicated in the pathogenesis of cutaneous fibrosis in scleroderma (10,15). We have previously demonstrated that, in the clinically involved skin of scleroderma patients with early active disease, Fli1-positive fibroblasts are either absent or only occasionally seen, whereas the majority of fibroblasts in healthy control skin express Fli1 at a relatively high level. Furthermore, there was an inverse correlation between the expression of Fli1 and type I collagen by dermal fibroblasts in lesional and healthy skin (10). Given the important biological role of Fli1, it is likely that its activity is regulated in a tissue-and context-dependent manner; however, to date very little is known about the posttranslational mechanisms regulating Fli1 function.
Because of the widespread role of Fli1 in various biological and pathological processes, it is important to gain additional insights into the molecular mechanisms governing the activity of Fli1. Here we showed that Fli1 is subjected to acetylation by the HAT activity of PCAF in vivo. We mapped the acetylation site to a unique lysine residue between EBS and the C-terminal activation domain at position 380 and provided convincing evidence that this residue is essential for DNA binding and transactivation of target genes. To our knowledge, this is the first report providing the principal mechanism responsible for the TGF-␤-induced dissociation of Fli1 from the collagen promoter.

MATERIALS AND METHODS
Reagents-Recombinant human TGF-␤1 was obtained from R&D Systems. The polyclonal rabbit anti-Fli1 antibody was described previously (7). The monoclonal mouse anti-Fli1 antibody was purchased form BD Bioscience. The monoclonal mouse and the polyclonal rabbit anti-acetylated lysine antibodies, anti-HA tag antibody, anti-phospho-Smad2 antibody, anti-Smad2 antibody, anti-phospho-ERK antibody, anti-ERK antibody, anti-phospho-Akt (Ser-473) antibody, and anti-Akt antibody were purchased from Cell Signaling Technology. Trichostatin A and antibodies for ␤-actin and FLAG tag were purchased from Sigma. Antibodies for PCAF, Sp1 and Ets1 were obtained from Santa Cruz Biotechnology. Antibody for CBP was purchased from Abcam. Antibody for calmodulin binding peptide was purchased form Millipore.
Cell Cultures-Human dermal fibroblast cultures were established from the foreskins of healthy newborns from the Medical University of South Carolina Hospital or from skin biopsies taken from the dorsal forearm of healthy donors in compliance with the Institutional Review Board for Human Studies as described previously (8,10). All studies used cells from passage number 3 to 6. Before stimulation with cytokines and infection with adenoviruses, fibroblasts were incubated in serum-free medium for 48 h. Human embryonic kidney 293T cells were purchased form ATCC and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum.
Plasmid Construction-pSG5-Fli1 was described previously (12). The coding sequence of Fli1 and Ets1 were cloned into pCTAP vector (Stratagene), which has two different tags, including a streptavidin-binding peptide and a calmodulinbinding peptide. Eighteen mutants of Fli1 gene were constructed by replacing lysine with arginine using pSG5-Fli1 and/or pCTAP-Fli1 as a template with the use of the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instruction. A Ϫ772 human ␣2(I) collagen (COL1A2)/CAT construct was generated as previously described (38). An expression vector of p300 was a gift from Dr. J. Boyes (Institute of Cancer Research, London, UK). Expression vector of CBP is a gift from Dr. R. Janknecht (Mayo Clinic, Rochester, MN). Expression vector of PCAF and PCAF/ ⌬HAT is a gift from Dr. T. Kouzarides (Cambridge University, Cambridge, UK). Expression vector of constitutive active TGF-␤ type I receptor (T␤RI T204D) is a gift from Dr. K. Miyazono (University of Tokyo, Tokyo, Japan). T␤RI T204D with mutations in the L45 loop (T␤RI T204D mL45) was generated by PCR-based mutagenesis as described previously (39). Plasmids were purified twice on cesium chloride gradients. At least two different plasmid preparations were used for each experiment.
Immunoblotting-Whole cell extracts were prepared using lysis buffer with the following contents: 1% Triton X-100, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 3 mM MgCl 2 , 1 mM CaCl 2 , proteinase inhibitor mixture (Roche Applied Science), 1 mM phenylmethylsulfonyl fluoride, and Trichostatin A (100 ng/ml). Protein extracts were subjected to SDS-PAGE and transferred to nitrocellulose membranes. Membranes were incubated overnight with primary antibody, washed, and incubated for 1 h with secondary antibody. After washing, visualization was performed by enhanced chemiluminescence (Pierce).
Immunoprecipitation-To precipitate non-tagged proteins, whole cell extracts (500 g) were pre-adsorbed with protein G-Sepharose beads (GE Healthcare) and incubated with 2 g of appropriate antibodies and then with protein G-Sepharose beads. Streptavidin-coupled agarose beads (Sigma) was used instead of protein G-Sepharose beads for immunoprecipitation of ectopically expressed tagged-Fli1. The precipitated proteins were subjected to immunoblotting.
RNA Isolation and Quantitative Reverse Transcription-PCR-RNA isolation and quantitative reverse transcription-PCR were performed as described previously (14). Briefly, 2 g of RNA isolated from cells using Tri reagent (MRC Inc) was reverse transcribed in 20 l of reaction volume using random primers and Transcriptor First Strand synthesis kit (Roche Applied Science). Real-time quantitative PCR was carried out using Sybr green master mix (Bio-Rad) on an Icycler machine (Bio-Rad) in triplicates. The sequences of Fli1 and COL1A2 primers were previously described (14). PCR conditions were 95°C for 3 min, followed by 40 cycles of 95°C for 30 s, 58°C for 1 min. Dissociation analysis for each primer pair and reaction was performed to verify specific amplification.
Reporter Gene Assay-Foreskin fibroblasts were grown to 50% confluence in 100-mm dishes, transfected with the indicated constructs along with pSV-␤-galactosidase using FuGENE6 (Roche Applied Science), and incubated for 48 h.
Extracts, normalized for protein content, were incubated with butyl-CoA and [ 14 C]chloramphenicol for 90 min at 37°C. Butylated chloramphenicol was extracted using an organic solvent (a 2:1 mixture of tetramethylpentadecane and xylene) and quantitated by scintillation counting.
In Vivo Acetylation Assay-293 T cells were transfected with expression vectors encoding tagged-Fli1 (0.1 g) and the indicated HAT protein (2 g), and incubated for 48 h. Whole cell lysates (500 g) were incubated with 10 l of streptavidin-coupled agarose beads overnight at 4°C. Precipitated proteins were subjected to immunoblotting using rabbit anti-acetylated lysine antibody. After the development, the membrane was stripped and reprobed with anti-calmodulin-binding peptide antibody to determine the total levels of ectopically expressed Fli1.
DNA Affinity Precipitation Assay-Two oligonucleotides containing biotin on the 5Ј-nucleotide of the sense strand were used. The sequences of these oligonucleotides are as follows: (i) COL1A2 EBS oligonucleotide, 5Ј-GAAAGGGCGGGG-GAGGGCGGGAGGATGCGGAGGGCGGAG-3Ј, which corresponds to bp Ϫ307 to Ϫ269 of the human ␣2(I) collagen promoter, containing both an EBS and GC-box, which can function as Sp1/3-binding sites, and (ii) COL1A2-EBS-M oligonucleotide, 5Ј-GAAAGGGCGGGGGAGGGCGGGAGTAT-GCGGAGGGCGGAG-3Ј, which has a mutated EBS of COL1A2-EBS oligonucleotide. These oligonucleotides were annealed to their respective complementary oligonucleotides, and double-stranded oligonucleotides were gel-purified and used. Whole cell extracts (500 g) prepared from 293T cells were incubated for 10 min at 4°C with gel shift binding buffer (10 mM Tris-HCl (pH 8.0), 40 mM KCl, 1 mM dithiothreitol, 6% glycerol, 0.05% Nonidet P-40), and 20 g of poly(dI-dC) (Pierce) in a final volume of 1 ml. Preclearing was performed by adding streptavidin-coupled agarose beads and incubating the mixture for 30 min with gentle rocking at 4°C. After centrifugation, the supernatant was incubated with 500 pmol of each double-stranded oligonucleotide overnight at 4°C with gentle rocking. Then 65 l of streptavidin-coated agarose beads was added, followed by a further 2-h incubation at 4°C. The protein-DNA-streptavidin-agarose complex was washed twice with Tris-EDTA (100 mM NaCl), twice with gel shift binding buffer, and once with phosphate-buffered saline. Precipitates were subjected to immunoblotting using the indicated antibodies.
Chromatin Immunoprecipitation Assay-The chromatin immunoprecipitation (ChIP) assay was carried out essentially as described previously (14). Briefly, cells were treated with 1% formaldehyde for 10 min. The cross-linked chromatin was then prepared and sonicated to an average size of 300 -500 bp. The DNA fragments were immunoprecipitated overnight with or without polyclonal anti-Fli1 antibody at 4°C. After reversal of cross-linking, the immunoprecipitated chromatin was amplified by PCR amplification of specific regions of the COL1A2 genomic locus. The primers were as follows: COL1A2/F-404, 5Ј-CTGGACAGCTCCTGCTTT-GAT-3Ј; COL1A2/R-233, 5Ј-CTTTCAAGGGGAAACTCT-GACTC-3Ј. The amplified DNA products were resolved by agarose gel electrophoresis.
Protein Stability Assays-Cycloheximide chase assays were performed as described previously with minor modification (40). 293T cells were cotransfected in 6-well plates with 0.1 g of wild-type Fli1 or Fli1 K380R plasmids and 2 g of PCAF or PCAF/⌬HAT plasmid. Twelve hours after transfection, the cells from three wells of a 6-well plate were pooled together and replated into five wells of 6-well plates to minimize the variation in transfection among the wells. Twelve hours after replating, 50 g/ml cycloheximide was added, and the cycloheximide treatment was terminated at 0-, 3-, 6-, 12-, and 24-h time points. 30 g of total protein from each sample was analyzed by immunoblotting.
Pulse-chase analysis was done as described previously (41). Briefly, for pulse labeling 293T cells were incubated in methionine-free minimal essential medium for 30 min and then pulsed in the same medium containing [ 35 S]methionine (1 mCi/ml) for 30 min. Pulse-labeled cells were chased for 0, 12, or 24 h in serum-free medium. Whole cell lysates (500 g) were subjected to immunoprecipitation using polyclonal anti-Fli1 antibody, followed by analysis of SDS-PAGE and autoradiography. The densities of bands were measured with a densitometer.
Pulse-chase experiments were analyzed by plotting the log of the densitometer volume for each band (with background from a blank lane on the gel subtracted) versus chase time and by obtaining a slope m for the plot by regression analysis. Turnover was assumed to be a first-order process, with t1 Adenoviral siRNA Vectors-An adenoviral vector expressing siRNA corresponding to PCAF gene (AdPCAFsiRNA) was generated. The following siRNA target sequence was used; 5Ј-TCGCCGTGAAGAAAGCGCA-3Ј (224 -242). Annealed double-stranded oligonucleotides were cloned into MluI and XhoI sites of pRNAT-H1.1 shuttle vector (Genescript). The shuttle vector with target siRNA sequence was linearized with PmeI and then electroporated into BJ5183-AD-1 (Stratagene) to generate recombinant AdEasy vector. The recombinant AdEasy plasmid after linearization with PacI enzyme was transfected into QBI-293 cells using Transfectin (Bio-Rad) for generation of adenovirus. The shuttle vector plasmid and AdEasy vector plasmid were sequenced to confirm the cloning. The primary adenoviral stock was then amplified and concentrated by cesium chloride density gradient centrifugation. Typical viral titer was 1 ϫ 10 10 pfu/ml. Control adenovirus vector was prepared in the same manner.
Statistical Analysis-Data presented as bar graphs are the means Ϯ S.D. of at least three independent experiments. Statistical analysis was performed using the Mann Whitney-U test (p Ͻ 0.05 was considered significant).

TGF-␤1 Stimulation Induces Acetylation of Fli1 and Dissociation of Fli1 from the ␣2(I) Collagen
Promoter in Human Dermal Fibroblasts-As an initial experiment, we investigated the effect of TGF-␤1 on the acetylation levels of Fli1. To this end, human adult dermal fibroblasts were stimulated with TGF-␤1, and acetylation levels of Fli1 were determined by immunoprecipitation with an anti-Fli1 antibody, followed by immunoblot-ting using an anti-acetylated lysine antibody. As shown in Fig.  1A, Fli1 was slightly acetylated in unstimulated cells. Upon TGF-␤1 treatment, acetylation levels of Fli1 were dramatically increased at 3 h. Of note, total levels of Fli1 were decreased in a time-dependent manner after TGF-␤1 stimulation, which is consistent with our previous report (10). The lower graph in Fig.  1A shows the acetylation levels of Fli1 normalized by its total levels. After TGF-␤1 stimulation Fli1 remains acetylated for up to 24 h. Thus, TGF-␤1 is a potent mediator of Fli1 acetylation. In contrast to protein levels, mRNA levels of Fli1 were not consistently affected by TGF-␤1 stimulation (Fig. 1B), suggesting that TGF-␤1 decreases the protein stability of Fli1.
We have previously demonstrated that Fli1 binds to human COL1A2 promoter in vitro and functions as its repressor (8). To further confirm these previous findings in vivo, we analyzed the effect of TGF-␤1 on the DNA binding status of Fli1 using ChIP analysis. As shown in Fig. 1C, Fli1 occupied the COL1A2 promoter in the absence of TGF-␤1 stimulation. Upon TGF-␤1 stimulation, Fli1 binding to DNA was decreased at 3 h and virtually disappeared at 24 h. These results suggest that TGF-␤1 mediates the dissociation of Fli1 from the COL1A2 promoter. Importantly, Fli1 acetylation and Fli1 dissociation from the COL1A2 promoter showed similar kinetics, suggesting that acetylation regulates the DNA binding status of Fli1.
PCAF Acetylates Fli1 in Vivo-To determine which HAT proteins mediate Fli1 acetylation, an in vivo acetylation assay was performed using 293T cells that express low levels of endogenous HAT proteins, including p300, CBP, and PCAF. Wild-type Fli1 with tags was coexpressed with p300, CBP, or PCAF in 293T cells, and immunoprecipitation was performed with streptavidin-coupled agarose beads. Ets1, another member of the Ets transcription factor family, had been shown previously to be acetylated in the presence of p300 (42) and therefore was included as a positive control. Overexpression of p300 or CBP did not affect the acetylation levels of Fli1 ( Fig. 2A, lanes  1 and 2), whereas acetylation of Ets1 was markedly induced ( Fig.  2A, lanes 6 and 7). In contrast, coexpression of PCAF resulted in a robust acetylation of Fli1 ( Fig. 2A, lane 3), but did not affect acetylation of Ets1 (data not shown). Furthermore, PCAF/ ⌬HAT, a deletion mutant that lacks HAT activity, failed to acetylate Fli1 (Fig. 2A, lane 4). To further prove this finding, immunoprecipitation with another anti-acetylated lysine antibody was performed using cell extracts prepared under the same condition. Consistently, ectopically expressed Fli1 was precipitated by an anti-acetylated lysine antibody only when coexpressed with PCAF (Fig. 2B, lane 3). We concluded from these experiments that acetylation of Fli1 is mediated by the HAT activity of PCAF.
Because previous reports demonstrated that HAT proteins interact with their acetylation targets, the interaction of Fli1 with HAT proteins was examined. Ectopically expressed Fli1 interacted with PCAF (Fig. 2C, lane 2), whereas the interaction of p300 with Fli1 was below the detectable level under regular exposure time (Fig. 2C, lane 1). In a longer exposure of the same blots, p300 was faintly detected (data not shown). This observation is consistent with the notion that Fli1 is preferentially acetylated by PCAF. Furthermore, the level of interaction between PCAF/⌬HAT and Fli1 was almost identical to that of PCAF and Fli1 (Fig. 2C, lanes 2 and 3), suggesting that loss of Fli1 acetylation by PCAF/⌬HAT is not due to the loss of interaction between these proteins. The effect of TGF-␤ on the FIGURE 1. TGF-␤1 induces acetylation of Fli1 and dissociation of Fli1 from the human ␣2(I) collagen promoter in human dermal fibroblasts. A, adult dermal fibroblasts were grown to subconfluence and serum-starved for 48 h, then treated with TGF-␤1 (2.5 ng/ml) for the indicated periods of time. Equal amounts of protein from each whole cell lysate were subjected to immunoprecipitation using mouse anti-Fli1 antibody, and acetylated forms of Fli1 were detected by immunoblotting using rabbit anti-acetylated lysine antibody (AcK). The same membrane was stripped and reprobed with rabbit anti-Fli1 antibody. B, Fli1 mRNA levels were determined in adult dermal fibroblasts using quantitative reverse transcription-PCR. The graph represents -fold change in Fli1 mRNA levels in response to TGF-␤1 treatment (2.5 ng/ml) in comparison to unstimulated controls, which were arbitrarily set at 100. C, chromatin was isolated from adult dermal fibroblasts and immunoprecipitated using rabbit anti-Fli1 antibody or beads alone (No-Ab). After isolation of bound DNA, PCR amplification was carried out using human ␣2(I) collagen (COL1A2) promoter-specific primers. Input DNA was taken from each sample before addition of an antibody. All bands show one representative of three independent experiments. NOVEMBER 30, 2007 • VOLUME 282 • NUMBER 48 interaction between Fli1 and PCAF was also examined in 293T cells (Fig. 2D). Minimum interaction was detected between ectopically expressed Fli1 and endogenous PCAF. As expected, upon TGF-␤1 stimulation, the interaction of Fli1 and PCAF was substantially increased. Taken together, these results indicate that PCAF interacts with Fli1 upon TGF-␤ stimulation and subsequently acetylates Fli1.

PCAF Acetylates Fli1
TGF-␤ Mediates Acetylation of Fli1 through a Smad-independent Pathway-Because Smad2 and -3 are the primary mediators of TGF-␤ signaling, we investigated if Smads are involved in the TGF-␤-dependent Fli1 acetylation. To this end, we used T␤RI T204D, a constitutively active TGF-␤ type I receptor, and T␤RI T204D mL45, which has a constitutively active kinase domain but lacks the ability to phosphorylate Smad2/3. These characteristics of T␤RI T204D mL45 were confirmed in Fig. 2E. ERK and Akt were phosphorylated by both T␤RI T204 and T␤RI T204D mL45, whereas Smad2 was phosphorylated by T␤RI T204D only. Acetylation levels of Fli1 were examined in the presence of T␤RI T204 or T␤RI T204D mL45. As shown in Fig. 2F, Fli1 was acetylated by PCAF, and acetylation levels were increased by coexpression of T␤RI T204D. Likewise, in the presence of T␤RI T204D mL45, Fli1 acetylation reached the levels comparable with T␤RI T204D. These results indicate that activation of the Smad2/3-signaling pathway is not required for the TGF-␤-dependent acetylation of Fli1.
Lysine 380 Is the Preferred Site for the PCAF-mediated Acetylation of Fli1-Fli1 contains a total of 24 lysine residues, any of which might be acetylated by PCAF. Therefore, to map the acetylation site(s), we replaced all lysine residues with arginines, either individually or in sets of two or three residues, to retain the positive charge while preventing acetylation. The Fli1 mutants were cotransfected with PCAF into 293T cells. The degree of acetylation of each Fli1 construct was assessed by an in vivo acetylation assay. Only the K380R mutation abolished acetylation of Fli1, whereas other mutations did not have a sig-

FIGURE 2. Fli1 is preferentially acetylated by HAT activity of PCAF in vivo.
A, 293T cells were transfected with wild-type Fli1 or Ets1 tagged with a streptavidin-binding peptide and a calmodulin-binding peptide along with the indicated HAT proteins, and incubated for 48 h. Total cell extracts were subjected to immunoprecipitation using streptavidin-coupled agarose beads (SA beads), followed by immunoblotting using rabbit anti-acetylated lysine antibody (AcK). To visualize the total levels of ectopically expressed Fli1, the same membrane was stripped and reprobed with the anti-calmodulin binding antibody. The levels of HAT proteins in cell lysates were determined by Western blotting. B, whole cell lysates of 293T cells were subjected to immunoprecipitation using mouse anti-acetylated lysine antibody followed by immunoblotting with anti-calmodulin-binding peptide antibody. The levels of HAT proteins, Fli1 and Ets1, in cell lysates were determined by Western blotting. C, tagged Fli1 was precipitated from whole cell extracts of 293T cells by streptavidin-coupled agarose beads. The amounts of the FLAG-tagged proteins and calmodulin-tagged proteins in the immunoprecipitates were determined by Western blotting. The input levels of each HAT protein were determined in whole cell extracts. D, 293T cells were transfected with expression vector encoding tagged wild-type Fli1 only and incubated for 48 h. Where indicated, cells were treated with TGF-␤1 (2.5 ng/ml) for the last 3 h. Whole cell extracts were incubated with streptavidin-coupled beads, and precipitates were subjected to immunoblotting using anti-PCAF antibody or anti-calmodulin-binding peptide antibody. E, 293T cells were transfected with T␤RI T204D or T␤RI T204D mL45 and incubated for 48 h. Whole cell lysates were subjected to immunoblotting with indicated antibodies. F, tagged-Fli1 was precipitated from whole cell extracts of 293T cells by streptavidin-coupled beads, followed by immunoblotting using a rabbit antiacetylated lysine antibody. Then, the same membrane was stripped and reprobed with the anti-calmodulin-binding peptide antibody. The levels of FLAG-tagged proteins and HA-tagged proteins in whole cell extracts were determined by immunoblotting. nificant effect on the degree of acetylation of Fli1 (Fig. 3A). We also performed immunoprecipitation using another anti-acetylated lysine antibody, followed by immunoblotting with an anticalmodulin-binding peptide antibody. As shown in Fig. 3B, this antibody failed to precipitate the Fli1 K380R mutant. Thus, loss of acetylation of Fli1 K380R mutant was confirmed using two distinct antibodies. There was no difference in the levels of interaction with PCAF between wild-type Fli1 and the Fli1 K380R mutant (data not shown), suggesting that decreased acetylation of Fli1 K380R mutant is not due to the loss of its interaction with PCAF. Taken together, these results indicate that lysine 380 appears to be the major site of PCAF-dependent acetylation in vivo. To further confirm this point, we compared the acetylation levels between the wild-type Fli1 coexpressed with PCAF/⌬HAT and the Fli1 K380R mutant coexpressed with PCAF. Upon longer exposure of the membrane, acetylated forms of Fli1 were detected under these conditions (Fig. 3C, lanes 2 and 3). The degree of acetylation of the Fli1 K380R mutant coexpressed with PCAF was comparable to that of the wild-type Fli1 coexpressed with PCAF/⌬HAT, indicating that Fli1 does not have any additional acetylation site besides lysine 380.
Acetylation Destabilizes Fli1 Protein-One of the functional consequences of acetylation is an alteration of protein stability. As shown in Fig. 1, decrease of total levels of Fli1 observed at 6 h after TGF-␤1 stimulation was preceded by Fli1 acetylation. These results suggested that TGF-␤ decreases the protein stability of Fli1 through PCAF-dependent acetylation. As an initial approach to exploring this possibility, we examined the effect of forced expression of PCAF on the steady-state levels of Fli1 and the Fli1 K380R mutant in 293T cells. As shown in Fig. 4A, PCAF decreased the steady-state levels of Fli1. In contrast, levels of the K380R mutant were not affected by PCAF. These results suggest that PCAF-dependent acetylation on lysine 380 decreases the protein stability of Fli1. To further strengthen this point, we carried out cycloheximide chase analysis using 293T cells. As shown in Fig. 4, B and C, the protein levels of wild-type Fli1 and K380R construct coexpressed with PCAF/ ⌬HAT showed similar kinetics. As expected, overexpression of PCAF led to a marked decrease in the stability of wild-type Fli1, whereas forced expression of PCAF exerted no effect on the stability of Fli1 K380R mutant. However, during a prolonged time course, cycloheximide chase analysis may not accurately reflect the degradation rate of target proteins under physiological conditions, because the synthesis of proteins regulating the protein degradation pathways could also be affected by cycloheximide. Therefore, we confirmed the results of cycloheximide treatment using pulse-chase analysis. As shown in Fig. 4D, both approaches produced similar results. To estimate a half-life of each Fli1 construct, the band density was measured by densitometry and plotted on a logarithmic scale graph (Fig.  4E). In the presence of PCAF/⌬HAT, the half-lives of wild-type Fli1 and K380R mutant were both ϳ10.39 h. The half-life of K380R mutant did not change significantly when coexpressed with PCAF. In contrast, overexpression of PCAF reduced the half-life of wild-type Fli1 to ϳ7.71 h. Taken together, these results support the conclusion that acetylation on lysine 380 decreases the protein stability of Fli1.
Acetylation of Fli1 Decreases Its DNA Binding Ability in Vitro and in Vivo-Another critical functional consequence of acetylation of transcription factors is a change in their DNA-binding ability. Therefore, we examined the ability of Fli1 to bind to the human COL1A2 promoter by DNA affinity precipitation assay. Untagged Fli1 constructs, either wild-type or K380R mutant, were transfected into 293T cells along with PCAF or PCAF/ ⌬HAT. Binding of each Fli1 construct to the COL1A2-EBS oligonucleotide (bp Ϫ307 to Ϫ269), containing EBS of the human COL1A2 promoter, previously characterized as a functional Ets binding site, was examined. Wild-type Fli1 coexpressed with PCAF/⌬HAT strongly bound to the oligonucleotide (Fig. 5A,  lane 2), whereas the DNA-binding ability of Fli1 coexpressed with PCAF was substantially decreased (Fig. 5A, lane 1). Wildtype Fli1 failed to bind to the mutated oligonucleotide (COL1A2 EBS-M) (Fig. 5A, lanes 3 and 4), confirming the specificity of Fli1 binding. In contrast, the Fli1 K380R mutant bound to the oligonucleotide much stronger than wild-type Fli1, and the binding levels were minimally affected by PCAF expression (Fig. 5B, lanes 3 and 4). These results indicate that PCAF-dependent acetylation inhibits the specific binding of Fli1 to the canonical DNA response element in vitro. We also examined the effect of PCAF on the DNA-binding ability of Sp1 and Ets1. As shown in Fig. 5C, in contrast to Fli1, PCAF overexpression did not affect binding levels of the endogenous Sp1 and Ets1 to  the COL1A2 EBS oligonucleotide, indicating that the regulation of DNA binding through PCAF-mediated acetylation is a distinct property of Fli1.
We next investigated whether PCAF-dependent acetylation affects Fli1 binding capacity to the COL1A2 promoter in vivo using ChIP analysis. Wild-type Fli1 or Fli1 K380R mutant were cotransfected with PCAF or PCAF/⌬HAT, along with the Ϫ772 COL1A2 promoter, into 293T cells. Cross-linked chromatin was immunoprecipitated with an antibody to Fli1, and the purified genomic DNA was amplified with primers specific to the COL1A2 promoter. The data in Fig.  5D provide clear evidence that Fli1 occupies the Ϫ404 to Ϫ233 region of the human COL1A2 promoter, but the binding is dramatically decreased in the presence of PCAF. In contrast, PCAF showed a minimum effect on the DNA binding status of the Fli1 K380R mutant. ChIP without antibody did not yield a significant amount of bound DNA (Fig. 5D, lanes 2, 5, 8, and 11). These data suggest that Fli1 is displaced from the COL1A2 promoter by PCAF-dependent acetylation in vivo.
PCAF-dependent Acetylation of Fli1 Decreases Its Inhibitory Effect on the Human COL1A2 Promoter Activity-To assess the impact of acetylation on Fli1 transactivation potential, we performed reporter gene analysis. Increasing amounts of the expression vector encoding PCAF were cotransfected together with fixed amounts of the Fli1 expression plasmid and the Ϫ772 COL1A2/CAT reporter gene. Transactivation was monitored by the CAT activity assay. As shown in Fig. 6A, consistent with previous results (8), TGF-␤1 significantly increased promoter activity of the Ϫ772 COL1A2 construct, and this effect was significantly suppressed by coexpression of wild-type Fli1. However, PCAF reversed the inhibitory effect of Fli1 on the TGF-␤1-mediated transactivation of the COL1A2 promoter in a dose-dependent fashion. Notably, coexpression of 0.5 g of PCAF almost completely abrogated the inhibitory effect of Fli1. To determine whether the effect of PCAF depends on Fli1 acetylation, we examined the ability of PCAF to reverse the Fli1 K380R mutant-mediated inhibition of COL1A2 promoter activity. As shown in Fig. 6B, the Fli1 K380R mutant exerted a greater inhibitory effect on the basal and the TGF-␤1mediated activity of the COL1A2 promoter than the wild-type Fli1. In contrast to the wild-type Fli1, PCAF overexpression was less efficient in reversing the inhibitory effect of the Fli1 K380R mutant on the TGF-␤1-mediated transactivation of the COL1A2 promoter. A partial rescue effect of PCAF may be attributed to its effect on the other components of the transcription complex involved in regulation of the COL1A2 promoter. As expected, PCAF/⌬HAT showed no effect on trans- Whole cell lysates were incubated with biotin-labeled oligonucleotides. Proteins bound to these nucleotides were isolated with streptavidin-coupled agarose beads, and Fli1, Ets1, and Sp1 were detected by immunoblotting. The levels of FLAG-tagged proteins and Fli1 in cell lysates were determined by Western blotting. D, 293T cells were transfected with the Ϫ772 human ␣2(I) collagen (COL1A2) promoter, along with the indicated expression vectors. After 48-h incubation, formaldehyde-cross-linked, moderately sheared chromatin was prepared. The DNA fragments were immunoprecipitated using rabbit anti-Fli1 antibody, and COL1A2 promoter fragments were detected using PCR.
activation mediated by the wild-type Fli1 and Fli1 K380R mutant. These results indicate that Fli1 loses its transactivation effect on the COL1A2 promoter by PCAF-dependent acetylation on lysine 380, which is consistent with results of the DNA affinity precipitation and ChIP analysis in 293T cells.
Gene Silencing of PCAF Abrogates TGF-␤1-mediated Fli1 Acetylation and Dissociation from the COL1A2 Promoter in Dermal Fibroblasts-The data presented so far have employed various overexpression systems. In our final set of experiments we asked whether endogenous PCAF has a similar impact on the function of endogenous Fli1. To this end, we employed an adenovirus-mediated RNA interference technique to knock down endogenous PCAF. We have established an adenoviral vector carrying a PCAF-specific siRNA sequence. Treatment of fibroblasts with 25 multiplicity of infection of AdPCAFsiRNA for 72 h resulted in a reduction of PCAF protein levels in human dermal fibroblasts by ϳ90% (Fig. 7A). Under this condition, we first examined the acetylation status of Fli1. As shown Fig. 7B, PCAF siRNA substantially decreased the TGF-␤-dependent acetylation of Fli1 and inhibited the TGF-␤-dependent decrease of steady-state levels of Fli1 protein. To confirm the DNA binding status of Fli1, ChIP analysis was also performed. As shown in Fig. 7C, TGF-␤1 decreased the association of Fli1 with the COL1A2 promoter at 6 h after stimulation. In contrast, TGF-␤1 had only a minimum effect on the levels of DNA-associated forms of Fli1 in AdPCAFsiRNA-transduced fibroblasts. Consistently, PCAF siRNA significantly decreased TGF-␤-induced mRNA levels of COL1A2 gene while showing only a marginal effect on its basal levels (Fig. 7D). These results strongly support the conclusion that PCAF-dependent acetylation regulates the dissociation of Fli1 from the COL1A2 promoter in dermal fibroblasts.

DISCUSSION
TGF-␤ plays a central role in the development of fibrotic diseases. Our previous studies have demonstrated that Fli1 is a repressor of the collagen type I gene and a potent inhibitor of the TGF-␤-induced profibrotic gene program (8,10,14). The focus of this study was to gain a better understanding of the molecular mechanisms whereby TGF-␤ abrogates the function of this repressor. We demonstrated that TGF-␤ elicits acetylation of Fli1 and the dissociation of Fli1 from the COL1A2 promoter. Fli1 is acetylated by the HAT activity of PCAF, and the major acetylation site is located at lysine 380. Furthermore, acetylation of Fli1 at this site decreases its protein stability and impairs its DNA-binding ability, leading to a decrease in its inhibitory effect on transcriptional activity of the COL1A2 gene. Importantly, we provide the evidence that siRNA-mediated PCAF silencing inhibits the TGF-␤-dependent acetylation of endogenous Fli1 and its subsequent dissociation from the endogenous COL1A2 promoter. Although it is formally possible that under different experimental conditions other lysine residues are also acetylated by PCAF, the functional analyses performed in this study indicate that acetylation at lysine 380 plays a critical role in the regulation of Fli1 transcription activity in the context of TGF-␤-mediated activation of dermal fibroblasts.
Eukaryotic transcription is a highly regulated process, and acetylation plays a critical role in this regulation. Lysine acetylation, which neutralizes a positive charge of the histone N termini, causes a reduction in the affinity of histone-DNA interactions, thereby destabilizing nucleosome structure and increasing accessibility of transcription factors to a genetic locus. In contrast, in the deacetylated form, histones are bound to DNA tightly, preventing transcription (16,17). Thus, in terms of chromatin condensation, acetylation is generally associated with transcriptional activation of target genes, whereas lack of acetylation tends to correlate with transcriptional repression, namely, two regulatory processes working in harmony to achieve appropriate levels of transcription. Accordingly, acetylation of transcription factors generally results in the transactivation of target genes (25,26,28,29,34,35,37), indicating that the acetylation event integrates chromatin condensation-and transcription factor-mediated regulation of gene transcription. Given that Fli1 is a potent repressor of collagen biosynthesis (8,10), the present observation that acetylation of Fli1 decreases its DNA-binding ability in the context of the COL1A2 promoter supports a canonical conceptual linkage between acetylation and gene transcription activation.
Only a single in vivo acetylation site was mapped on lysine 380, located between the DNA binding domain and the C-ter-minal activation domain of Fli1. This finding is consistent with the notion that HAT proteins generally modify very few lysines within a given protein (16,17). However, similarly to other Ets transcription factors, ER81 and PU.1 (25,26), acetylated lysine 380 in human Fli1 protein is not a part of the HAT consensus recognition motif, GKXXP (43,44).
The effect of acetylation on DNA-binding capacity appears to depend on the location of the modified site within the protein. In most cases, acetylation of lysine residues within the DNA-binding domain has an inhibitory effect on DNA binding (YY1, p65) (34,45), because the negative charge of the acetyl group neutralizes the positive charge of the lysine that is responsible for tightening the bonds between the transcriptional complex and the promoter (16,17). On the other hand, acetylation in a domain nearby the DNA-binding domain generally results in increased DNA binding (p53, E2F1, MyoD, GATA-1, ER81, and PU.1) (25-29, 32, 33, 36). In contrast, acetylation of Fli1 on lysine 380, in a region adjacent to the DNA binding domain, results in disruption of its DNA binding. A similar case was reported for the HMGI(Y), a nonhistone nuclear protein, in which the acetylation site is located directly adjacent to the DNA binding domain, and acetylation decreases DNA binding (46). A reasonable interpretation of the present observation might be that acetylation on lysine 380 results in a conformational change of Fli1, preventing the ETS domain from binding to DNA. Further studies are required to elucidate fully the exact mechanisms involved.
Although numerous reports in the last decade have established the relationship between acetylation and protein stability, acetylation generally enhances the half-life of target proteins (25,28,29,47). This is because ubiquitin-mediated proteasomal degradation is often prevented by competition between acetylation and ubiquitination occurring on the same lysine residues. So far, very little is known about the negative effect of acetylation on the stability of target proteins. In the present study, however, acetylation on lysine 380 was found to decrease the half-life of Fli1 protein. A similar observation was reported for the hypoxia-inducible factor-1␣. ARD1-mediated FIGURE 7. Gene silencing of PCAF inhibits the effect of TGF-␤ on Fli1 acetylation and DNA binding ability in vivo. A, adult dermal fibroblasts were transduced with 25 multiplicity of infection of adenovirus PCAFsiRNA (AdPCAFsiRNA) or non-silencing siRNA (NS) for 72 h. Whole cell extracts were subjected to immunoblotting for PCAF. B, under the same condition, cells were treated with TGF-␤1 (2.5 ng/ml) for the indicated period of time. Acetylation levels of Fli1 were determined as described in the legend of Fig. 1A. The expression levels of PCAF in whole cell extracts were determined by immunoblotting. C, under the same condition, formaldehyde-crosslinked, moderately sheared chromatin was prepared. In some experiments, cells were stimulated with TGF-␤1 (2.5 ng/ml) for the last 6 h. The DNA fragments were immunoprecipitated using rabbit anti-Fli1 antibody, and the presence of the human ␣2(I) collagen (COL1A2) promoter fragments were detected using PCR. D, under the same condition, cells were treated with TGF-␤1 (2.5 ng/ml) for the last 24 h. mRNA levels of COL1A2 gene were determined using quantitative reverse transcription-PCR. The graph represents -fold change in COL1A2 mRNA levels in comparison to unstimulated and non-silencing siRNA adenovirus-transduced controls, which were arbitrarily set at 100. The mean Ϯ S.D. from three separate experiments are shown. *, p Ͻ 0.05 versus nonsilencing siRNA adenovirus-transduced control cells stimulated with TGF-␤1. acetylation enhances the interaction of hypoxia-inducible factor-1␣ with an ubiquitin ligase, pVHL, and a subsequent ubiquitination of hypoxia-inducible factor-1␣ leads to a proteasomal degradation of hypoxia-inducible factor-1␣ (48). A similar mechanism may be applicable to Fli1, because we observed a reversal of the PCAF-induced Fli1 protein degradation in 293T cells by proteasomal inhibitor MG132 (data not shown). The specific mechanisms involved in the regulation of Fli1 protein stability are under investigation in our laboratory.
Recently, there is accumulating evidence for the critical role of acetylation in the TGF-␤-mediated transactivation of target genes. There are several reports that showed positive influence of acetylation on Smad2/3 signaling. For example, p300/CBPand PCAF-mediated acetylation of Smad2 on lysine 19 located in the MH1 domain increased its DNA-binding ability through a conformational change and subsequently resulted in transactivation of target genes (49). Another group showed that Smad3 is acetylated on lysine 378, which is located in the MH2 domain, by p300/CBP and PCAF (50). Although the effect of acetylation at this position on Smad3 DNA binding was not shown, the authors demonstrated that acetylation of Smad3 on lysine 378 has a stimulatory effect on the expression of several TGF-␤target genes. In a separate study, ectopic expression of PCAF increased the TGF-␤-dependent DNA binding of Smad3, although the acetylation of Smad3 was not detected (51). Based on these recent findings and the results obtained in our study, we propose that acetylation is the critical post-translational modification whereby TGF-␤ coordinately regulates DNAbinding activity of transcriptional activators, such as Smad3 and repressors such as Fli1, resulting in increased extracellular matrix production.
A recent report demonstrated the association between enhanced type I collagen expression and epigenetic repression of the Fli1 gene in scleroderma fibroblasts (15). This finding is consistent with our previous observation that constitutive down-regulation of Fli1 contributes to the pathogenesis of cutaneous fibrosis in scleroderma (10). Significantly, we have found that acetylation levels of Fli1 are increased in a subset of lesional scleroderma fibroblast cell lines compared with healthy control fibroblasts. 4 Thus, acetylation-related posttranslational modification of Fli1 may be caused by epigenetic factors involved in the pathogenesis of cutaneous fibrosis in scleroderma.