Site-specific acetylation of the fetal globin activator NF-E4 prevents its ubiquitination and regulates its interaction with the histone deacetylase, HDAC1.

Acetylation provides one mechanism by which the functional diversity of individual transcription factors can be expanded. This is valuable in the setting of complex multigene loci that are regulated by a limited number of proteins, such as the human beta-globin locus. We have studied the role of acetylation in the regulation of the transcription factor NF-E4, a component of a protein complex that facilitates the preferential expression of the human gamma-globin genes in fetal erythroid cells. We have shown that NF-E4 interacts directly with, and serves as a substrate for, the acetyltransferase co-activator PCAF. Acetylation of NF-E4 is restricted to a single residue (Lys(43)) in the amino-terminal domain of the protein and results in two important functional consequences. Acetylation of NF-E4 prolongs the protein half-life by preventing ubiquitin-mediated degradation. This stabilization is PCAF-dependent, since enforced expression in fetal/erythroid cells of a mutant form of PCAF lacking the histone acetyltransferase domain (PCAFDeltaHAT) decreases NF-E4 stability. Acetylation of Lys(43) also reduces the interaction between NF-E4 and HDAC1, potentially maximizing the activating ability of the factor at the gamma-promoter. These results provide further demonstration that co-activators, such as PCAF, can influence individual transcription factor properties at multiple levels to alter their effects on gene expression.

Proteins with intrinsic histone acetyltransferase (HAT) 1 activity function as co-activators of transcription through direct acetylation of specific lysine residues within the N-terminal tails of core histones (1). This modification leads to destabilization of local nucleosomal structure with induction of an open chromatin configuration, allowing access of the transcriptional machinery to core promoters (reviewed in Refs. [2][3][4]. The covalent modification of histones by acetylation also provides recognition sites for factors involved in gene activation or repression (5). Two of the most widely studied protein families with acetylase activity are the GCN5/PCAF (1,6) and p300/ CBP proteins (7,8). The substrates for these factors are far more diverse than the histone proteins alone and include transcription factors such as p53 (9), GATA-1 (10,11), erythroid Kruppel-like factor (EKLF) (12), SCL (13), E2F1 (14), transcriptional co-regulators (15), DNA-binding proteins (16), retroviral proteins (17), and nonnuclear proteins (18). The sequelae of acetylation of these proteins range from altered DNA binding or cellular localization to changes in protein stability or protein-protein interactions (reviewed in Ref. 19). These significant changes in function occur in the context of modification of a small number of lysine residues within an individual protein and thus provide mechanisms to expand the roles for single factors in the regulation of complex multigenic loci.
One such locus is the ␤-like globin cluster in which a linear array of five genes is expressed in a highly regulated tissuespecific and developmentally specific manner (20). To date, very few transcription factors that specifically influence the temporal profile of globin gene expression have been identified. One of these, EKLF is a red cell-specific activator that is critical for switching on high level adult ␤-globin expression (21)(22)(23). EKLF binds to the CACCC element in the proximal ␤-promoter, leading to chromatin remodeling and transcriptional activation (24 -28). The transcriptional activity of EKLF is regulated by site-specific acetylation by p300 and CBP that results in an enhanced activation potential of the protein and increased binding to the SWI/SNF chromatin remodeling complex (12,29).
NF-E4 is another globin-specific transcription factor (30). This protein forms the stage selector protein complex (SSP) with the ubiquitous transcription factor CP2 (31). The SSP facilitates the interaction of the ␥-globin genes with the powerful enhancer elements contained in the locus control region in fetal erythroid cells through binding to the stage selector element (SSE) in the proximal ␥-promoter (32). Enforced expression of NF-E4 in the human fetal/erythroid K562 cell line and human cord blood progenitors leads to increased fetal globin gene expression (30). In cord blood progenitors, where active competition between fetal and adult globin genes occurs, enforced expression of NF-E4 also leads to a reduction in ␤-globin gene expression (30).
In this study, we demonstrate that NF-E4 is a direct target of the co-activator PCAF. The resultant acetylation of Lys-43 in NF-E4 results in both an increase in the stability of the protein due to diminished targeting by ubiquitination and an alteration in protein-protein interactions that favor transcriptional activation. Thus, a single amino acid modification results in complementary functional changes that enhance the ability of NF-E4 to positively regulate fetal globin gene expression.

EXPERIMENTAL PROCEDURES
Cell Culture-K562 cells were grown in RPMI medium 1640 supplemented with 10% fetal bovine serum. 293T cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. All cells were grown at 37°C and in 5% CO 2 supplemented with 50 units of penicillin/ml and 50 g of streptomycin/ml.
Stable K562 cell lines overexpressing HA-NF-E4, HA-K43R, PCAF, and PCAF⌬HAT were generated according to the protocols described previously, except that only the top 10% of GFP-positive cells were collected (30). For the co-immunoprecipitation experiment, 293T cells were transiently transfected with various plasmids by a calcium phosphate method (Invitrogen). For the reporter gene assays, native K562 cells or stable K562 cell lines overexpressing HA-NF-E4 and HA-K43R were co-transfected with HS2-53␥-Luciferase, containing hypersensitivity site 2 of the ␤-globin locus control region linked to the Ϫ53 ␥-promoter relative to the CAP site and the firefly luciferase reporter gene (32), and pCH110, containing the ␤-galactosidase gene driven by SV40 promoter as an internal transfection efficiency control by electroporation. For the TSA induction experiments, the transfected cells were divided into two aliquots, one of which was cultured in various concentrations of TSA for 20 h and the other in standard medium alone. For all other transfection experiments, cells were harvested after 36 h and lysed, and luciferase activity was determined on a Monolight 2001 luminometer using the luciferase assay kit (Promega). ␤-Galactosidase activity was measured in a spectrophotometer using the enzyme assay system (Promega), and the relative luciferase activities were calculated by dividing the luciferase activity by the ␤-galactosidase activity to correct for transfection efficiency. Three independent transfections were performed in duplicate.
Immunoprecipitation and Immunoblotting-Cells were lysed in icecold lysis buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 1% Nonidet P-40, 10 mM sodium butyrate) containing a protease inhibitor mixture (Roche Applied Science) and cleared by centrifugation. Immunoprecipitations were carried out by adding appropriate antibodies plus protein G-Sepharose beads, followed by incubation at 4°C. The immunoprecipitates were washed extensively, subjected to SDS-PAGE, and transferred to nitrocellulose membranes. The membranes were incubated with various specific antibodies and then washed extensively prior to incubation with peroxidase-conjugated anti-rabbit or antimouse immunoglobulin G. After further extensive washes, the blots were visualized by using ECL reagents (Amersham Biosciences). All immunoprecipitations were performed in duplicate.
Recombinant Protein Expression and GST Pull-down Assay-GSTfusion proteins were produced in BL21 Escherichia coli as described previously (30). 35 S-Labeled NF-E4 and PCAF synthesized using the T7 TNT kit (Promega) and Tran 35 S-label (ICN) were incubated with GST fusion proteins prebound to glutathione beads at 4°C overnight. The samples were washed extensively and subjected to SDS-PAGE. The gels were dried and analyzed by autoradiography.
Determination of Protein Half-life-K562 cells stably expressing HA-NF-E4, HA-K43R, PCAF, and PCAF⌬HAT were treated with cycloheximide (100 M) to stop protein synthesis and incubated for the indicated times. Cells were lysed, and lysates were subjected to SDS-PAGE and Western blotting probed with the indicated specific antibodies. The protein signals were scanned and quantitated by PhosphorImager analysis software (Amersham Biosciences).
In Vitro Protein Acetylation Assay-2-3 g of the indicated GST fusion protein or SCL and 200 ng of acetyltransferase protein were incubated in a reaction containing 50 mM Tris-HCl, pH 8.0, 10% glycerol, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 mM sodium butyrate, and 30 M acetyl-coenzyme A or 1 l of [ 14 C]acetyl-CoA for 1 h at 30°C. The reaction mixture was subjected to SDS-PAGE and analyzed by autoradiography of dried gels or by Western blotting with anti-acetyl lysine antibody.
In Vitro Ubiquitination Assay-Recombinant protein was used as the substrate in a ubiquitin reaction containing 20 mM Hepes, pH 7.5, 5 mM MgCl 2 , 2 mM dithiothreitol, 2 mM ATP, 5 g of ubiquitin, 20 M MG132, and 5 l of crude rabbit reticulocyte (Promega) for 1 h at 30°C. In some cases, GST fusion proteins were acetylated prior to the ubiquitination reaction by PCAF in the presence or absence of unlabeled acetyl-CoA. The samples were subjected to SDS-PAGE and transferred to nitrocellulose membranes followed by determination of ubiquitination by Western blotting analysis.
In Vivo Ubiquitination Assay-293T cells were transfected with pCDNA3.1-FLAG-ubiquitin, and cell lysate was prepared 36 h later. For immunoprecipitation, 1 mg of protein was incubated with ␣-HA antibody at 4°C overnight before protein G beads were added for 2 h. The beads were washed twice with NaCl (1 M) in Tris-buffered saline supplemented with Nonidet P-40 (1%), ␤-mercaptoethanol (0.05%), and EDTA (1 mM). Proteins were loaded onto SDS-PAGE followed by immunoblot analysis with the indicated antibodies and enhanced chemiluminescence detection.
Electrophoretic Mobility Shift Assay-The electrophoretic mobility shift assay was performed as previously described (31) using recombinant proteins. In some cases, recombinant NF-E4 was acetylated in vitro prior to binding DNA. Specific antibodies against NF-E4 or CP2 or specific or nonspecific cold oligonucleotide competitors in molar excess were added to the reaction mixture for 30 min on ice before the addition of the labeled SSE oligonucleotides.

Acetylation of NF-E4 in Vitro and in
Vivo-Distinct functional parallels exist between the globin regulatory transcription factors EKLF and NF-E4. Both factors are expressed throughout erythroid ontogeny and yet exert their predominant effects at distinct developmental time points. Acetylation of EKLF is critical for its role as an adult globin gene activator, and we postulated that a similar post-translational modification might also regulate NF-E4 function. We therefore examined the ability of NF-E4 to serve as a substrate for two acetyltransferases, PCAF and CBP. A purified GST-NF-E4 fusion protein was incubated with recombinant CBP or PCAF in the presence of [ 14  To determine whether NF-E4 was acetylated in vivo in a fetal/erythroid environment, human K562 cells were pulse-labeled with either [ 35 S]methionine/cysteine or sodium [ 3 H]acetate in the presence of TSA. Cellular extract was subjected to immunoprecipitation with anti-NF-E4 antibody, and precipitates were washed under stringent conditions. As a control, extracts were also immunoprecipitated with preimmune sera. As shown in Fig. 1B, NF-E4-specific antisera detected both the 35 S-labeled NF-E4 (left panel) and acetylated NF-E4 (right panel). Neither species was detected with preimmune sera. To directly examine the role of PCAF in the acetylation of NF-E4 in a cellular context, we co-transfected the human cell line 293T with mammalian expression vectors containing a hemagglutinin (HA) epitope-tagged NF-E4 (HA-NF-E4), and either the wild-type PCAF cDNA (PCAF) tagged with a FLAG epitope or a mutant PCAF that lacked the histone acetyltransferase domain (PCAF⌬HAT) tagged with a FLAG epitope, or vector alone (13,14). Cellular extracts from the transfected cells were immunoprecipitated with preimmune sera (lane 1) or anti-HA antibody (lanes 2-4) and immunoblotted with either anti-HA antibody (␣-HA) or antibody specific for acetylated lysine (␣-AcK) (Fig. 1C). An increase in the level of acetylated NF-E4 was observed in the presence of PCAF compared with vector alone (compare lanes 2 and 3). The GST were incubated with [ 14 C]acetyl-CoA and bacterially expressed PCAF or CBP as indicated. Reaction products were fractionated by SDS-PAGE, and the dried gel was exposed to x-ray film. B, human K562 cells were pulselabeled with either [ 35 S]methionine/cysteine or [ 3 H]acetate in the presence of 50 nM TSA and lysed. Lysates were subjected to immunoprecipitation (IP) with NF-E4 antibody or preimmune sera (PI), and the resulting immunoprecipitates were fractionated by SDS-PAGE and analyzed for incorporation of [ 35 S]methionine/cysteine or [ 3 H]acetate by autoradiography. C, 293T cells were transfected with expression vectors for HA-tagged NF-E4 and FLAG-tagged PCAF or PCAF that lacked the histone acetyltransferase domain (PCAF⌬HAT). Cellular extracts from the transfected cells were immunoprecipitated with preimmune sera or anti-HA antibody and immunoblotted (IB) with either anti-HA antibody (␣-HA) or antibody specific for acetylated lysine (␣-AcK). Extracts were also directly immunoblotted with anti-FLAG antibody, which detects both PCAF species (␣-Flag).
level of acetylated NF-E4 was decreased in cells transfected with PCAF⌬HAT, consistent with its dominant negative role (lane 4). The expression of HA-NF-E4 and FLAG-PCAF/ PCAF⌬HAT were similar in the transfected lines (lower panels). These findings indicate that NF-E4 serves as a substrate for the acetyltransferase, PCAF.
NF-E4 Interacts with PCAF in Vitro and in Vivo-Acetylation of transcription factors is usually mediated by a direct interaction between the factor and the specific acetyltransferase. To determine whether NF-E4 interacted directly with PCAF, we performed glutathione S-transferase (GST)-chromatography assays ( Fig. 2A). In vitro transcribed/translated [ 35 S]methionine-labeled NF-E4 (lane 1) was applied to glutathione-Sepharose beads adsorbed with GST alone (lane 2); GST-PCAF-(352-832) (lane 3), which contains key regulatory domains including the HAT domain; and a positive control, GST-CP2 (lane 4). The labeled protein was retained on both the GST-PCAF and GST-CP2 matrices but not on GST alone. To further localize the region of NF-E4 that interacted with PCAF, we generated a series of truncation deletions of NF-E4 as GST fusion proteins and examined their ability to retain [ 35 S]methionine-labeled PCAF (Fig. 2B). Only NF-E4 fusion proteins containing the first 25 amino acids were capable of interacting with PCAF (lanes 3-5). In contrast, fusion proteins that lacked only the N-terminal 17 amino acids did not interact (lane 6), and no PCAF was retained on the GST alone matrix (lane 2). To confirm the interaction in a cellular context, we transfected 293T cells with mammalian expression vectors containing the NF-E4 cDNA tagged with a hemagglutinin epitope (HA-NF-E4) and the PCAF cDNA tagged with a FLAG epitope (FLAG-PCAF) (Fig. 2C). Cellular extract was immunoprecipitated with an unrelated antibody (mock) or anti-FLAG antibody and immunoblotted with either anti-HA antibody (top panel) or anti-FLAG antibody (bottom panel). Both PCAF and NF-E4 FIG. 2. NF-E4 and PCAF interact in vitro and in cellular extracts. A, purified GST and GST fusion proteins containing amino acids 352-832 of PCAF (GST-PCAF-(352-832)) or full-length CP2 (GST-CP2) preadsorbed to glutathione-Sepharose beads were incubated with 35 S-labeled in vitro transcribed/translated (IVT) NF-E4. Specifically bound protein was eluted from washed beads and visualized by autoradiography after SDS-PAGE. Input represents 20% of the in vitro translated NF-E4 used in the assay. Each experiment was performed in duplicate. B, purified GST and GST fusion proteins containing amino acids 1-25, 1-48, 49 -179, and 17-179 (top panel) of NF-E4 preadsorbed to glutathione-Sepharose beads were incubated with 35 S-labeled PCAF. Specifically bound protein was eluted from washed beads and visualized by autoradiography (bottom panel) after SDS-PAGE. Input represents 5% of the in vitro translated PCAF used in the assay. C, 293T cells were transfected with expression vectors for HA-tagged NF-E4 and FLAG-tagged PCAF. Cellular extracts were immunoprecipitated (IP) with antibody to the FLAG epitope (␣-Flag) or an unrelated antibody (mock) and analyzed by immunoblotting (IB) with anti-HA and anti-PCAF antibodies. NF-E4 protein was identified only in precipitates using anti-FLAG antisera. D, K562 cell extract was immunoprecipitated with either anti-NF-E4 or anti-PCAF antibodies. Preimmune sera served as the control. Immunoprecipitates were fractionated by SDS-PAGE and immunoblotted with anti-PCAF antibody. PCAF protein was detected in precipitates from both anti-NF-E4 and anti-PCAF antibodies but not preimmune sera.
were immunoprecipitated with the anti-FLAG antibody, but not the unrelated antibody, indicating that the two proteins interact in a cellular context. Immunoprecipitation of lysate from nontransfected cells with the anti-FLAG antibody failed to bring down NF-E4 (data not shown). We then examined whether this interaction occurred in the absence of enforced expression of the proteins. K562 cell extract was immunoprecipitated with either preimmune sera or antibody to NF-E4 or PCAF, and the precipitates were electrophoresed and immunoblotted with anti-PCAF antibody. As shown in Fig. 2D, PCAF co-immunoprecipitated with NF-E4 (lane 2), indicating that these two factors form a complex in native cells. No PCAF was observed with preimmune sera (lane 1).
Identifying Acetylation Sites in NF-E4 -To identify the residue(s) in NF-E4 that are acetylated by PCAF, we utilized our series of GST-NF-E4 truncation mutants (Fig. 3A, left panel). The proteins were subjected to a PCAF-dependent in vitro acetylation assay using unlabeled acetyl-CoA, and protein acetylation was detected using an antibody specific for acetylated lysine (␣-AcK) (Fig. 3A, top right panel).  (Fig. 3A, bottom right panel). This acetylation pattern was duplicated when acetylated lysine was detected by incorporation of [ 14 C]acetyl-CoA in the in vitro assay (data not shown). These results localized the site of acetylation of NF-E4 to the residues between 26 and 48 and indicated that the acetylation status of the NF-E4 protein was accurately reflected by Western analysis with the antibody specific for acetylated lysine.
Acetylation of proteins involves the attachment of an acetyl group to the side chain of lysine residues. Examination of the predicted amino acid sequence of NF-E4 in this region revealed a single potential site for NF-E4 acetylation at lysine 43. To determine whether this residue was acetylated by PCAF in vitro, we generated a full-length NF-E4 protein in which the lysine at position 43 was altered to arginine (K43R) (Fig. 3B,  top panel). Cleaved GST fusion proteins of the full-length wildtype NF-E4 and K43R mutant were incubated with PCAF in the presence of acetyl-CoA and sodium butyrate, separated by SDS-PAGE, and immunoblotted with anti-acetyl lysine antibody or anti-NF-E4 antibody. As shown in Fig. 3B (bottom  panel), despite producing abundant mutant and wild-type proteins as detected with anti-NF-E4 antibody, no acetylation was detected in the setting of the K43R mutation. This mutation did not affect the ability of PCAF to bind to NF-E4 or the nuclear localization of the protein (data not shown). To examine the effect of the K43R mutation on NF-E4 acetylation in vivo, we transfected 293T cells with mammalian expression vectors containing the HA-tagged NF-E4 cDNA (HA-NF-E4) or a tagged NF-E4 cDNA containing the K43R mutation (HA-K43R). Extract from both lines was immunoprecipitated with ␣-HA antisera and immunoblotted with either ␣-AcK or ␣-HA. As shown in Fig. 3C, despite abundant expression of the mutant and wild type NF-E4 proteins as demonstrated with anti-HA antibody (␣-HA), only a faint band was detected in the setting of the K43R mutation (␣-AcK). In contrast, the wildtype protein displayed prominent acetylation consistent with Lys-43 acting as the sole site of in vitro acetylation and the major site of in vivo PCAF-dependent acetylation.

NF-E4 DNA Binding Is Not Affected by Acetylation in Vitro-
The DNA-binding properties of many transcription factors are altered by acetylation (19). To examine the effects of this modification on NF-E4 DNA binding, we performed an electrophoretic mobility shift assay with recombinant CP2 and NF-E4 and an SSE probe from the human ␥-promoter (Fig. 4). Under the conditions used, recombinant CP2 failed to bind the SSE probe in isolation (lane 1). Recombinant NF-E4 was preincubated with PCAF in the presence and absence of acetyl-CoA using the conditions defined in Fig. 1A. Both forms of NF-E4 failed to bind the SSE in isolation (data not shown). When increasing amounts of unacetylated NF-E4 (lanes 2-4) or acetylated NF-E4 (lanes 5-7) were combined with rCP2 and probe, a robust fast migrating complex was observed. The formation of this complex was unaffected by the acetylation status of the rNF-E4 but was altered by specific antibodies to both CP2 and NF-E4 (lanes 8 and 9). A weaker, slower migrating complex that was similarly unaffected by the acetylation status of NF-E4 was also noted. We attributed this to a higher order NF-E4⅐CP2 complex (30), since it was dependent on the amount of NF-E4 added and was supershifted with both antibodies. To further confirm the specificity of the protein-DNA complexes, we examined the effects of specific and nonspecific unlabeled oligonucleotides. As shown in lanes 11-13, the presence of a 10 -100-fold molar excess of unlabeled SSE probe resulted in diminution/loss of the SSP complexes. In contrast, a 100-fold excess of an unrelated unlabeled oligonucleotide did not alter the SSP⅐SSE complex (lane 14). These findings indicate that acetylation of NF-E4 does not alter its DNA binding properties.
Acetylation Stabilizes NF-E4 Protein by Inhibiting Its Ubiquitination-Acetylation has recently been shown to increase the stability of a number of diverse transcription factors (14,33,34). We examined the influence of acetylation on NF-E4 protein stability by determining the half-life of endogenous NF-E4 in K562 cells stably transfected with retroviral vectors carrying wild-type PCAF or PCAF⌬HAT or the empty vector as a control (Fig. 5A). Cellular extracts were prepared at various time points after cycloheximide treatment and subjected to SDS-PAGE and immunoblotted with antibodies to NF-E4 or tubulin as a control. The expression of PCAF and PCAF⌬HAT was shown to be comparable in the two cell lines by immunoblotting extracts with ␣-PCAF (top right panel). As shown in data from a representative experiment (left panel) and graphically from all experiments (bottom right panel), an increase in endogenous NF-E4 stability was observed in the cells transfected with wild-type PCAF compared with vector alone. In contrast, the half-life of NF-E4 in K562 cells transfected with FIG. 4. Acetylation of NF-E4 does not affect DNA binding. Recombinant NF-E4 and CP2 were used in electrophoretic mobility shift assays as indicated with a 32 P-labeled probe containing the SSE sequence from the proximal ␥-promoter. For lanes 2-7, increasing amounts of NF-E4 were incubated with PCAF in the presence or absence of acetyl-CoA prior to the addition of rCP2 and the electrophoretic mobility shift assays. Acetylation of the recombinant protein was confirmed by immunoblotting with anti-AcK antibody (data not shown). The CP2⅐NF-E4 complexes were disrupted with anti-NF-E4 and anti-CP2 antibodies (lanes 8 and 9) and excess unlabeled SSE probe (lanes [11][12][13]. No effect was observed with an unrelated unlabeled probe (lane 14). The SSP complex is shown, and the band marked with the asterisk is a higher order CP2⅐NF-E4 complex that is altered by both antibodies and specific unlabeled probe.
PCAF⌬HAT was significantly reduced, suggesting that the stability of NF-E4 is dependent on PCAF-mediated acetylation of Lys-43.
We therefore compared the stability of wild type and acetylation-deficient (K43R) NF-E4. K562 cells were stably transfected with retroviral vectors carrying HA-tagged wild-type (HA-NF-E4) or K43R mutant NF-E4 (HA-K43R), and extracts were processed as described above (Fig. 5B, left panel). The half-life of the transfected, tagged HA-NF-E4 was less than 1 h (right panel), considerably shorter than the endogenous NF-E4 (Fig. 5A). We attributed this to levels of endogenous PCAF expression that were limiting in the context of a marked increase in NF-E4 expression in the transfected cells. Surprisingly, the half-life of the acetylation-deficient HA-K43R was almost 3 h (right panel), 3 times longer than the wild type protein. The levels and stability of the control protein, tubulin, were comparable in extracts from both transduced lines. These results suggested that acetylation of lysine 43 plays an important role in the stabilization of the NF-E4 protein in cells. However, they also suggest that lysine 43 may also play a role in the targeted degradation of NF-E4.
The enhanced stability of both acetylated NF-E4 and the acetylation-deficient K43R mutant was reminiscent of the findings reported with the transcription factors Smad7 and the SREBP family (33,34). In these proteins, the sites of acetylation that enhance protein stabilization are also the lysine residues targeted by ubiquitination. Consequently, either acetylation or mutation of these lysine residues increases protein stability. To determine whether acetylation of NF-E4 altered the susceptibility of the protein to ubiquitination, we utilized recombinant NF-E4 in both acetylated and nonacetylated forms as substrates in reconstituted in vitro ubiquitination assays (Fig. 5C, top panel). Nonacetylated NF-E4 was ubiquitinated in the presence of rabbit reticulocyte lysate (RRL) (lane 3) and migrated as a high molecular weight polyubiquitinated species (NF-E4-Ub n ). Acetylation of NF-E4 markedly reduced its ubiquitination (compare lanes 2 and 3).  (right panel, top). To ensure that equal amounts of protein were loaded in each well, the levels of tubulin in the samples were estimated by Western blotting using anti-tubulin antibodies (␣-Tub). The amount of NF-E4 at each time point is plotted as a percentage of the amount at the start of the chase normalized to the tubulin levels and represents the average Ϯ S.E. of the mean of three independent experiments (right panel). B, K562 cells were stably transfected with vectors containing HA-tagged wild type or K43R acetylation-deficient NF-E4. Cells were processed as in A, and a representative experiment is shown (left panel). The amount of HA-NF-E4 or HA-K43R at each time point is plotted as a percentage of the amount at the start of the chase normalized to tubulin and represents the average Ϯ the S.E. of the mean of three independent experiments (right panel). C, GST-HA-NF-E4 was used as the substrate in in vitro ubiquitination reactions in the presence or absence of rabbit reticulocyte lysate (RRL). Where indicated, GST-HA-NF-E4 was acetylated prior to the ubiquitination reaction (lane 2). The samples were resolved by SDS-PAGE, and the ubiquitination of NF-E4 was determined with anti-ubiquitin antibodies (␣-Ub). The polyubiquitinated NF-E4 is marked as NF-E4-Ub n . The amount of HA-tagged NF-E4 and its acetylation were determined with anti-HA (␣-HA) and anti-acetyl lysine (␣-AcK) antibodies, respectively. D, 293T cells were stably transfected with an empty vector or vectors containing HA-tagged wild type or K43R acetylation-deficient NF-E4. These lines were subsequently transfected with a mammalian expression vector containing FLAG-tagged ubiquitin or the vector alone, as a control. Thirty-six hours after transfection, cell lysates were immunoprecipitated (IP) with anti-HA antibodies (␣-HA) and resolved by SDS-PAGE. The ubiquitination of the wild type and mutant NF-E4 was determined by Western blotting (IB) using anti-ubiquitin antibodies (␣-Ub) or anti-FLAG antibodies (␣-Flag). In the top panel, the asterisk denotes a nonspecific band, and no ubiquitinated species are seen in this panel. In the second panel, the polyubiquitinated NF-E4 species are marked as NF-E4-Ub n , and the immunoglobulin heavy chain is also shown. The amounts of HA-tagged wild type and mutant protein were determined in the cell lysates by Western blotting with anti-HA antibodies (third panel), and the expression of FLAG-tagged ubiquitin is shown in the bottom panel.
examine the effect of mutation of Lys-43 on the ubiquitination of NF-E4 in a cellular context, we utilized the 293T cells stably transfected with retroviral vectors containing the HA-tagged wild-type (HA-NF-E4) or K43R mutant NF-E4 (HA-K43R) or the empty vector control (Fig. 5D). The levels of expression of the tagged NF-E4 forms were similar in both lines, with a small increase in the K43R protein level reflecting its increased stability (shown in ␣-HA, third panel). These lines were then transfected with a mammalian expression vector containing FLAG-tagged ubiquitin (ϩFlag-Ub, second panel) or the empty vector (w/o Flag-Ub, top panel). The levels of expression of the FLAG-Ub were equivalent in all three lines (bottom panel). Cellular extracts were prepared and immunoprecipitated with anti-HA antibody prior to immunoblotting with either antiubiquitin or anti-FLAG antibody. In the absence of FLAG-Ub expression (w/o Flag-Ub, top panel), the ␣-HA-immunoprecipitate from all three lines contained no ubiquitinated NF-E4 species, with only a nonspecific band (asterisk) visualized in all three lanes. We attributed this to low levels of endogenous ubiquitin expression in these lines. When FLAG-Ub was expressed (ϩFlag-Ub, second panel), high molecular weight polyubiquitinated NF-E4 species (NF-E4-Ub n ) were seen in the immunoprecipitates from the wild type NF-E4-expressing cells. These polyubiquitinated proteins were not seen in cells containing the empty vector, and more significantly, they were not seen in the K43R-expressing cells. These results suggest that lysine 43 is also targeted by ubiquitination in vivo.
Effect of K43R Mutation on NF-E4 Transcriptional Activity-We predicted that functional effects of stabilization of NF-E4 would be reflected by an increase in the activity of the ␥-promoter. To test this, we utilized the stable K562 cells in which expression of wild type or K43R mutant NF-E4 had been enforced and a vector-transduced line as a control. These lines were then transiently transfected with a plasmid containing the ␥-promoter truncated to Ϫ53 relative to the CAP site linked to a luciferase reporter gene (32). This construct contains the SSE and the TATA box but none of the other known upstream regulatory elements and is therefore specifically responsive to NF-E4. A ␤-galactosidase-expressing plasmid was also used to monitor transfection efficiency. As shown in Fig. 6A, a greater than 2-fold increase in ␥-promoter activity was observed in cells expressing the wild-type NF-E4. Surprisingly, the increase in ␥-promoter activity observed in the K43R-expressing cells was significantly less than in the wild-type cells despite the higher level of expression of the mutant protein (Fig. 6B). This finding suggested that stabilization of the NF-E4 protein was insufficient in isolation to enhance the protein's activation potential and that acetylation of the protein must play an additional role in NF-E4-mediated ␥-gene activation.
To examine this further, we investigated the effects of the histone deacetylase inhibitor TSA on ␥-promoter activity in a transient transfection assay. K562 cells were co-transfected with the ␥-promoter-luciferase reporter and the ␤-galactosidase-expressing plasmid and divided into three flasks, two of which were induced with TSA in varying concentrations. As shown in Fig. 6C, a dose-dependent increase in ␥-promoter activity was seen with TSA induction. This finding was consistent with the enhancement of endogenous ␥-gene expression seen with this compound (35,36).
Diminished Binding of Acetylated NF-E4 to HDAC1-Altered protein-protein interactions are a well established consequence of transcription factor acetylation (19). The affinity of EKLF for the SWI/SNF chromatin-remodeling complex is enhanced by this modification (29), prompting us to consider whether acetylation of NF-E4 could influence its protein interactions to enhance its activation potential. We had previously demonstrated that NF-E4 and the histone deacetylase HDAC1 were co-immunoprecipitated from extract derived from K562 cells using anti-NF-E4 antibody (Fig. 7A). HDAC1 is a member FIG. 6. Acetylated NF-E4 activates a minimal ␥-globin promoter. A, K562 cells were co-transfected with an empty vector or vectors containing HA-tagged wild type or the acetylation-deficient K43R NF-E4 mutant and a human ␥-globin promoter reporter construct containing the Ϫ53 ␥-promoter (relative to the CAP site) linked to a firefly luciferase reporter gene. A ␤-galactosidase-expressing vector was used to monitor transfection efficiency. After 36 h, cell lysates were prepared and assayed for luciferase and ␤-galactosidase activity as described under "Experimental Procedures." The results represent the average Ϯ S.E. of the mean of three independent experiments performed in duplicate. B, the levels of HA-NF-E4 and HA-K43R were determined in the cell lysates by Western blotting (IB) using anti-HA antibodies (␣-HA). The levels of tubulin in the samples were estimated by Western blotting using anti-tubulin antibodies (␣-Tub). C, K562 cells were transfected with the human ␥-globin promoter fragment linked to a firefly luciferase reporter gene. A ␤-galactosidase-expressing vector was used to monitor transfection efficiency. After the transfection, the cells were divided into three and grown in the presence and absence of the stated concentrations of TSA for 20 h. Cell lysates were then prepared and assayed for luciferase and ␤-galactosidase activity as described under "Experimental Procedures." The results represent the average Ϯ S.E. of three independent experiments performed in duplicate. of a family of deacetylases involved in transcriptional repression and gene silencing (reviewed in Ref. 37). To determine whether acetylation of NF-E4 altered its affinity for HDAC1, we utilized recombinant GST-HA-NF-E4 in both acetylated and nonacetylated forms and a GST alone control in pull-down experiments with K562 cell extract (Fig. 7B). Acetylation of the recombinant protein was confirmed by immunoblotting with anti-AcK antibody (bottom panel). Analysis of the eluate from the three GST matrices with anti-HDAC1 antibody (␣-HDAC1) demonstrated that nonacetylated NF-E4 bound significantly greater amounts of HDAC1 derived from K562 extract than acetylated NF-E4. In contrast, CP2 from K562 cell extract was retained equally on both acetylated and nonacetylated NF-E4 as detected by CP2 antibody (␣-CP2). GST alone bound neither protein. These findings suggest that acetylation of NF-E4 plays a dual role in enhancing the activation potential of the factor via increased protein stability and diminished association with the transcriptional repressor, HDAC1.

DISCUSSION
Our previous studies have suggested that NF-E4 plays a role in regulation of the human ␤-globin locus through activation of the ␥-genes and competitive silencing of ␤-gene expression (30). The mechanisms underlying these activities are gradually emerging, and appear to hinge on both critical protein/protein interactions and post-translational modification of the factor. In the fetal erythroid environment, NF-E4 is a component of a multiprotein complex that contains the transcriptional activator CP2 (30,31). We now provide evidence that the acetyltransferase PCAF also complexes with NF-E4 in this environment and covalently modifies the protein. This modification has a number of sequelae, including increased stability of the protein through resistance to ubiquitination and alteration of protein partners to favor interactions that promote transcriptional activation.
The PCAF-interacting domain of NF-E4 is contained within the amino-terminal region of the NF-E4 protein. Although the full-length NF-E4 protein contains a number of lysine residues, it appears that in the context of PCAF, only the single residue at position 43 is acetylated. We found no evidence that CBP directly modifies the NF-E4 protein. In this regard, NF-E4 is unlike several of the other erythroid transcription factors that bind to the globin locus, including GATA-1, EKLF, and p45 NF-E2, all of which are acetylated by CBP (10 -12, 38). Interestingly, these factors have all been directly implicated in recruitment of acetyltransferases to the adult ␤-globin promoter (29,39,40). In contrast, the fetal globin regulatory factors, NF-E4 and FKLF-2, are both acetylated and functionally modified by PCAF, raising the possibility that stage-specific recruitment of acetyltransferases could occur at the globin locus (41). This could provide a potential explanation for the paradox of the highly stage-specific effects of factors such as EKLF and NF-E4 that are observed despite expression of these factors at all stages of erythroid development.
The functional consequences of acetylation of several of the globin locus binding factors have been examined previously. Modification of the MafG component of NF-E2 augments the DNA binding activity of this complex (38). Enhanced DNA binding has also been reported for acetylated GATA-1 (10), although there is some conjecture about this conclusion (11). We observed no effect of acetylation on the DNA-binding properties of NF-E4, a finding that parallels the lack of effect of acetylation on EKLF binding to its cognate site (29). The effects of acetylation on EKLF involve enhanced interaction between the factor and the SWI/SNF chromatin-remodeling complex (29). This results in maximization of the activation potential of EKLF at the ␤-promoter. The acetylation of NF-E4 also appears to maximize its activation potential through increased levels of the protein and a diminished association with the transcriptional repressor HDAC1. HDAC1 is a member of a family of proteins involved in mammalian gene silencing through deacetylation of core histones (37). Several HDAC-dependent co-repressor complexes have been identified in mammalian cells including mSin3A (42)(43)(44)(45)(46)(47) and NuRD/Mi2 (48,49). Although the nature and role of such complexes are yet to be established at the ␥-promoter, our identification of HDAC1 and more recently another NuRD component, SHARP (50) 2 as protein partners of NF-E4 suggest the potential for an NF-E4-dependent repressor complex to assemble on the promoter. As such, it appears that NF-E4 may play a dual role in fetal globin gene regulation depending on the developmental context.
Modification of a number of transcription factors has been shown to influence their interaction with HDAC1. The phosphorylation status of NF-B determines whether it associates with CBP/p300 or HDAC1 (51). The acetylation status of MyoD and YY1 also influences the ability of these factors to bind HDAC1 (52,53). In both cases, acetylation increases the interaction with the deacetylase, in contrast to the reduced association of acetylated NF-E4 with HDAC1 that we observed. An-2 S. M. Jane and J. M. Cunningham, unpublished observations. FIG. 7. Reduced binding of HDAC1 to acetylated NF-E4. A, K562 cell extract was immunoprecipitated with anti-NF-E4 antibodies (␣-NF-E4). Preimmune sera served as the control (PI). Immunoprecipitates (IP) were fractionated by SDS-PAGE and immunoblotted (IB) with anti-HDAC1 antibody (␣-HDAC1). HDAC1 protein was detected in precipitates from anti-NF-E4 antibody but not preimmune sera. B, GST-HA-NF-E4 coupled to glutathione-Sepharose beads was preincubated with PCAF in the presence and absence of acetyl-CoA (Ac-CoA) prior to the addition of K562 cell extract. GST alone served as the control. Specifically bound proteins were eluted from washed beads and resolved by SDS-PAGE. The presence of CP2 and HDAC1 in the eluates was detected by Western blotting with their respective antisera (␣-CP2 and ␣-HDAC1). The level of HA-NF-E4 on the beads and the acetylation status of bound NF-E4 in each experiment were determined by Western blotting using anti-HA (␣-HA) and anti-acetyl lysine (␣-AcK) antibodies, respectively. The input of CP2 and HDAC1 is shown on the left. Each experiment was performed in triplicate. other hematopoietic transcription factor, SCL, has also been shown to have an impaired interaction with the mSin3A repressor complex including HDAC1 when acetylated (13). This may provide a novel mechanism to expand the potential roles of a factor at a single complex locus.
The enhanced stability of NF-E4 we observed in the context of acetylation has been reported for a number of transcription factors, including E2F1, members of the SREBP family, and Smad7 (14,33,34). In the latter two examples, it was demonstrated that specific lysine residues were targeted by ubiquitination and that acetylation of these residues prevented the subsequent ubiquitination of the protein. Our data indicate that a similar mechanism governs NF-E4 stability. The potential link between acetylation status of proteins and their ubiquitination has been strengthened by the identification of ubiquitin pathway components in the HDAC6 complex (54). NF-E4 degradation may be dependent on its deacetylation and ubiquitination, and it is tempting to speculate that this may be mediated by NF-E4 serving as a substrate for HDAC1. Our future studies will explore this possibility.