Acetylation of EKLF Is Essential for Epigenetic Modification and Transcriptional Activation of the {beta}-Globin Locus

Posttranslational modifications of transcription factors providealternate protein interaction platforms that lead to varieddownstream effects. We have investigated how the acetylationof EKLF plays a role in its ability to alter the β-likeglobin locus chromatin structure and activate transcriptionof the adult β-globin gene. By establishing an EKLF-nullerythroid line whose closed β-locus chromatin structureand silent β-globin gene status can be rescued by retroviralinfection of EKLF, we demonstrate the importance of EKLF acetylationat lysine 288 in the recruitment of CBP to the locus, modificationof histone H3, occupancy by EKLF, opening of the chromatin structure,and transcription of adult β-globin. We also find thatEKLF helps to coordinate this process by the specific associationof its zinc finger domain with the histone H3 amino terminus.Although EKLF interacts equally well with H3.1 and H3.3, wefind that only H3.3 is enriched at the adult β-globin promoter.These data emphasize the critical nature of lysine acetylationin transcription factor activity and enable us to propose amodel of how modified EKLF integrates coactivators, chromatinremodelers, and nucleosomal components to alter epigenetic chromatinstructure and stimulate transcription.

INTRODUCTION

Top

INTRODUCTION

The mammalian β-like globin locus contains related genesthat are expressed at different times during development. Thisdevelopmentally regulated pattern of β-globin gene expressionis known as globin gene switching (9, 68). In humans, the firstgene expressed in the cluster is ${varepsilon}$ -globin in the yolk sac, followedby a switch in expression to ${gamma}$ -globin in the fetal liver. Thesecond switch is then to β-globin within the bone marrow.In addition to their individual promoters, the expression ofthese globin genes is controlled by a far-upstream region calledthe locus control region (LCR), a region whose chromatin exhibitstissue-specific DNase hypersensitivity and boundary elementsand whose subsequences behave as transcriptional enhancers (17).The regulation of this large β-globin cluster is thoughtto occur in part by competition of each globin member promoterfor interaction with the LCR that, in addition to an endogenoussilencing mechanism, enables the expression of only one memberat the right time in development. Over the years, a number oferythroid enriched proteins, such as GATA-1, EKLF (erythroidKrüppel-like factor), and NF-E2, have been identified andstudied for their ability to bind to the globin promoter sequencesand DNase hypersensitive sites (HSs) at the LCR (23, 44, 45).EKLF is an erythroid-specific transcription factor that activatesadult β-globin expression by binding to the CACCC elementin the promoter (49).

Until recently, the biological evidence for EKLF's functionhad been limited to its role in activating β-globin geneexpression. For example, EKLF^–/– murine embryosdie at day 14.5 of gestation due to a lethal anemia that resultsfrom the failure to produce adult β-globin (beta major)transcripts at the time of the switch to definitive erythropoiesis(54, 59). Further study also revealed a specific loss of developmentallyrelevant DNase I-accessible chromatin in the proximal β-globinpromoter and at HS3 at the β-LCR (70, 74). Since the degreeof DNase I hypersensitivity of a given locus correlates withnucleosomal remodeling and chromatin accessibility, these findingssuggested that EKLF plays a role in these processes at the β-globinlocus in definitive erythroid cells.

However, a paradox that had emerged from the results of theexpression and genetic studies came from the observation thatEKLF's message is also expressed in primitive (yolk sac) erythroidcells and multipotential hematopoietic cell lines (67), in spiteof the knockout data suggesting that it is functionally requiredonly for definitive erythropoiesis. In any case, a number ofrecent observations have begun to resolve this picture. First,the results of microarray analyses, followed by direct testing,indicate that there are a range of EKLF transcriptional activationtargets in addition to adult β-globin that fall withinheme-biosynthetic, erythroid membrane, and hemoglobin-interactingprotein categories (20, 24, 27, 60). Second, the results ofcareful phenotypic analyses reveal that EKLF-null primitiveerythroid cells are not normal (20). Third, the results of promoter(40) and prospective hematopoietic-cell analyses (22) indicatethat EKLF is expressed prior to erythroid commitment, particularlywithin the megakaryocyte-erythroid progenitor. Fourth, EKLFinteracts with corepressors, such as Sin3A, histone deacetylase1, and Mi2β (12, 66). Finally, EKLF inhibits megakaryopoiesiswhile at the same time promoting erythroid maturation and differentiation(22, 66).

Because the EKLF protein is expressed in primitive and definitiveerythroid cell populations, as well as in the megakaryocyte-erythroidprogenitor, posttranslational modifications and the resultantchanges in protein-protein interactions are likely criticalfor generating functionally altered EKLF states at differentstages of embryonic development and/or hematopoietic differentiation.EKLF undergoes multiple modifications, including phosphorylation(56), sumoylation (66), ubiquitination (61), and acetylation(76, 77), and some of these have been shown to alter EKLF protein-proteininteractions (12, 66, 77). Of particular relevance to the presentstudy, the acetylation of EKLF occurs subsequent to its interactionwith the acetyltransferases CBP and P300 at lysine residues288 and 302 near its zinc finger domain (77). The acetylationof EKLF at 288K is critical for the optimal transactivationof β-globin in transient assays (77), and the results ofin vitro studies show that this modification enhances its abilityto recruit the large erythroid complex (ERC-1) that containsSWI/SNF chromatin-remodeling proteins to the promoter (4), likelyvia its interaction with BRG1 (10, 33, 34). The general conclusionsabout EKLF function gleaned from the results of the molecularstudies have been supported in vivo, where the results of chromosomeconformation capture assays have established that EKLF is crucialfor the formation of the active chromatin hub at the β-locus(19). These observations are further supported by the resultsof studies showing that ablation of EKLF leads to the absenceof CBP, BRG1, TBP, and RNA pol II at the β-globin promoterin vivo (7). As a result, EKLF plays roles both in establishingthe correct three-dimensional structure at the β-like globinlocus and in activating the transcription of the adult β-globingene. Although the results of transient assays and in vitroanalyses are highly suggestive, it has not been verified thatEKLF posttranslational modifications, particularly acetylation,are critical for these in vivo observation, at the endogenousβ-like globin locus. In addition, the role of EKLF modificationin its ability to recruit coactivators, open chromatin, andinduce transcription in vivo has not been addressed.

For the present studies, we have developed a rescue system withina newly established EKLF-null erythroid cell line to addressthese issues. These have led to novel results regarding theimportance of EKLF acetylation in these processes and to someunanticipated interactions with nucleosomal components thatenable us to present a more-detailed model of EKLF's mode ofaction at the β-like globin locus.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

EKLF^–/– cell line and manipulations. The 18SCF cell line was established from EKLF-null fetal liversafter infection with the J2 retrovirus as described previously(14). Briefly, embryonic day 12.5 (E12.5) EKLF^–/–p53^–/– fetal liver cells were cultured at a densityof 5 x 10⁵ cells/ml in the presence of 20% supernatant obtainedfrom the culture of 2-J2 retroviral producer cells in I score'smodified Dulbecco's medium (Gibco) with 10% fetal calf serum,20% pokeweed mitogen-spleen conditional medium (21), 250 ng/mlof stem cell factor (SCF) (Gibco), 0.3 U/ml of erythropoietin(Sigma-Aldrich), 1,000 U/ml of interleukin-3 (Gibco), and 1%penicillin-streptomycin (Sigma-Aldrich). After 72 h of fetalliver cell culture, many dividing cells formed "grapes" in suspension.These grapes of cells were picked and cultured individually.After expansion in the same medium, 10 out of 30 potential clonessurvived, and among these, two clones had characteristics oferythroid cells (cellular morphology and capacity to express ${alpha}$ -globin).

The cell surface marker expression in actively growing 18SCFcells was monitored after staining with the following antibodies(Pharmingen): CD19-peridinin chlorophyll protein, B220-fluoresceinisothiocyanate (FITC), CD3e-FITC, CD4-phycoerythrin (PE), CD41-FITC,CD11b-PE, Ly6G-FITC, CD71-PE, Ter119-PE, CD117-allophycocyanin,CD34-PE, and Ly6A/E-FITC. The cells were analyzed by using FACSCalibur(Becton Dickinson) and FlowJo (TreeStar). All gates were drawnbased on negative controls for each sample. Cell morphologywas assessed by examining the cells under a 100x oil immersionobjective after a cytospin and May-Grunwald Giemsa staining(Sigma).

Full-length wild-type or acetylation mutant EKLF was clonedinto the EcoRI site of the murine stem cell virus-internal ribosomeentry site-green fluorescent protein (GFP) (MIG) retroviralvector (72). After calcium phosphate cotransfection with thepCL-Eco packaging vector (53) into 293T cells, the supernatantwas used to infect 18SCF cells as described previously (26).The GFP positivities correlated with EKLF mRNA and protein expressionlevels, and all experiments used clones that exhibited similarlevels of wild-type or acetylation mutant protein (as in Fig.1).

View larger version (40K):
[in this window]
[in a new window]

FIG. 1. EKLF rescue system in 18SCF cells. (A) A diagram of the EKLF MIG retrovirus used to infect EKLF-null 18SCF cells is shown (top) along with the results of FACS analyses indicating the scatter and typical level of GFP-positive cells after infection. LTR, long terminal repeat; IRES, internal ribosome entry site; hGFP, humanized GFP; SSC, side scatter; FSC, forward scatter; R5, cells sorted for GFP analysis, R6, GFP-nonexpressing cells; R7, GFP-expressing cells. (B) Time course of GFP loss after infection. Actively growing cells were reanalyzed by FACS for GFP expression at various time points (days, as indicated) after their initial selection at day 1 (GFP_O+) and in comparison to cells not expressing GFP (GFP_O–). (C) Total RNA from E13.5 fetal liver cells, uninfected 18SCF cells, or rescued (GFP-positive [GFP+]) cells was analyzed for the expression of EKLF or adult β-major globin (β-maj) by semiquantitative RT-PCR. Hypoxanthine phosphoribosyltransferase (HPRT) served as a positive control.

Cell culture and stable line analyses. 293T or 745A murine erythroleukemic (MEL) cells were maintainedin Dulbecco's modified Eagle's medium containing 10% serum andglutamine. MEL cell lines with short hairpin EKLF (shEKLF) weregifts from F. Morle (8). Plasmids containing FLAG-H3.1 (kindgift from K. Wilson [51]) or FLAG-H3.3 (kind gift from Y. Nakatani[69]) were transfected into 293T or MEL cells by using FuGene6 (Roche) or Tfx-50 reagent (Promega), respectively, accordingto the manufacturer's instructions. Stable clones in MEL celllines were selected by increasing exposure to G418, expanded,and used for further experiments. RNA extraction and quantitativeRT-PCR were performed as described previously (66).

DNase hypersensitivity assay. Nuclei were prepared from 18SCF cells or infected and GFP-selectedprogeny carrying wild-type EKLF, 288K/R, 302K/R, or 288/302/K/Ras described previously (57), followed by quantitation of theHS formation as described previously (46). DNase I digestionconditions were selected after extensive preliminary experimentsto determine the correct enzyme concentration and time of incubation.

ChIP. Chromatin immunoprecipitation (ChIP) was performed as describedpreviously (28, 40), with the following modifications. Briefly,2 x 10⁶ cells were cross-linked in 1% formaldehyde at room temperature.The cells were sonicated, lysates were cleared by centrifugation,10% was removed to serve as input, and the remaining materialwas incubated with the desired antibodies. Chromatin occupancylevels were quantified by real-time PCR measuring Sybr green(Qiagen) fluorescence relative to a titrated chromatin inputstandard curve. Primers were designed by using Primer Expresssoftware (Applied Biosystems), and their specificities wereverified by showing that each primer pair generated a singleamplicon and dissociation curve. The results were read withan ABI Prism 7900HT sequence detection system and analyzed withABI software. Primer sequences are available on request.

In vitro pull-down assay. Core histones H2A, H2B, H3.3, and H4 (kind gifts from N. Mermod[3]); glutathione S-transferase (GST)-tagged H3.3, H3.1, orvarious truncation mutants (50); and EKLF constructs were preparedas described previously (76). GST-tagged H3.3 or H3.1 or varioustruncation mutants (50) and EKLF constructs were used as describedpreviously (76). Briefly, bacterial cells harboring the GSTconstructs were induced with isopropyl-β-D-thiogalactopyranoside(IPTG). To isolate the overexpressed proteins, the cells werelysed by sonication in incubation buffer containing 50 mM Tris-HCl(pH 7.5), 200 mM NaCl, 0.1% NP-40, and protease inhibitors,cleared by centrifugation, and passed through a GST-Sepharosecolumn. The column was washed with incubation buffer. The purifiedprotein retained in the column was separated on sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels (Bio-Rad)and detected by Coomassie staining. For the pull downs, equalamounts of GST-fusion proteins coupled to glutathione Sepharosebeads (final concentration, approximately 50 nM) were incubatedwith 10 µl in vitro-translated [³⁵S]methionine protein(STP3 T7 kit; Novagen) overnight at 4°C in a total of 1ml of incubation buffer. The beads were washed five times withincubation buffer containing 300 mM and 0.2% NP-40, boiled inSDS loading buffer, and electrophoresed on SDS-PAGE gels. Allgels were exposed to EN³HANCE solution (NEN Life Science product),followed by detection by autoradiography.

RESULTS

Top

RESULTS

Establishment of EKLF^–/– erythroid lines. To facilitate our ability to monitor the in vivo effects ofmutant EKLF proteins at endogenous loci, we generated an erythroidcell line that is deficient in EKLF and used these immortalizedcells for rescue experiments. E13.5 fetal liver cells from EKLF-nullembryos (59) were immortalized by infection with the J2 retrovirus(14) and expanded from clonally derived colonies. The selectedclone, 18SCF, grows well in erythropoietin plus SCF. The resultsof fluorescence-activated cell sorter (FACS) analyses indicatethat these cells do not express B-cell, T-cell, or megakaryocyticmarkers; that they exhibit some macrophage/granulocyte expression;and that they express very high levels of CD71 but very lowlevels of Ter119 (Table 1). They also express moderate levelsof c-kit, CD34, and Sca-1. Morphological observations supportthese characteristics. As expected, EKLF is not expressed andβ-globin levels are not detectable by semiquantitativereverse transcription-PCR (RT-PCR) (not shown). We have expandedthe 18SCF line for up to 8 weeks for the present studies.

View this table:
[in this window]
[in a new window]

TABLE 1. Cell surface characteristics of 18SCF cells^a

Reintroduction of EKLF and rescue of endogenous β-globin expression. We cloned wild-type EKLF and its derivatives containing mutationsin each of its acetylation sites into a MIG vector (72) andprepared viral supernatants. We initially tested whether the18SCF cells could be rescued by retroviral infection with theMIG EKLF retrovirus. We found that infection efficiency wasquite low ( $~$ 1%) with the use of spinoculation or retronectinapproaches with high-titer virus (Fig. 1A). However, infectedcells were sorted via the GFP marker, and these could be expandedfor analysis for 2 to 3 weeks, after which the positive, EKLF-expressingpopulation began to decrease (Fig. 1B). Using this material,we found that β-globin expression is only observed in theEKLF-expressing cells (Fig. 1C), thus providing us with a suitablerescue system.

We next addressed whether the acetylation mutants could rescueβ-globin expression as successfully as wild-type EKLF.Extracts from each of the infected and sorted 18SCF populationsexpress similar levels of EKLF protein and quantified message(Fig. 2A and B). We found that while mutation of 302K had aminimal effect on β-globin rescue, mutation of 288K (orthe double mutant at 288/302/K/R) produced an essentially inactiveEKLF (Fig. 2C). These data show that the availability of 288K,but not of 302K, for acetylation is critical for the optimaltransactivation ability of EKLF at the endogenous β-globinpromoter. This supports the earlier observations derived fromtransient assays suggesting that the acetylation of lysine 288is crucial for EKLF activity (76) but that the acetylation oflysine 302 is not critical for EKLF's function as an activator(12).

View larger version (17K):
[in this window]
[in a new window]

FIG. 2. EKLF rescues adult β-globin expression. (A) The upper panel shows the results of a Western blot analysis with anti-EKLF antibody of the protein extracts from 18SCF cell lines not infected (–, lane 1) or infected with wild-type (WT, lane 2) or acetylation mutant EKLF (288K/R, lane 3; 302K/R, lane 4; or 288/302/K/R, lane 5). The bottom panel is the loading control. (B and C) Results of quantitative RT-PCR analysis of EKLF (B) and adult β-globin RNA (C) in 18SCF cell lines not infected or infected with wild-type or mutant EKLF. The bars correspond to the lanes in panel A. The experiment was performed in triplicate, and the expression values were normalized to those for glyceraldehyde-3-phosphate dehydrogenase. Error bars show standard deviations. In all cases, the cells were sorted for GFP expression after infection. After subsequent expansion, the GFP-positive cells were subjected to Western blotting and quantitative RT-PCR analysis.

DNA binding, coactivator recruitment, and formation of an open chromatin structure are dependent on acetylation of EKLF lysine 288. The mouse LCR consists of a series of tissue-specific DNaseI HSs that have enhancer and possibly insulator or boundaryelement properties in the erythroid lineage. The LCR directsthe transcription of four downstream β-globin genes thatare arranged in the order of their developmental expression(2). Our earlier studies had not shown any qualitative differencesin DNA binding in vitro by any of the EKLF acetylation mutantsto the CACCC sequence located at the –90 region of theβ-major promoter (77). Using the rescue system with 18SCFcells, we tested whether the wild type and the acetylation mutantsof EKLF can bind to the β-major promoter equally well invivo by using ChIP. We found that wild-type EKLF and the 302K/Rmutant are enriched at the promoter (Fig. 3A). However, occupancyby the 288K/R mutant and the 288/302/K/R double mutant at theβ-globin promoter is virtually undetectable (Fig. 3A).These results demonstrate that the availability of lysine 288for acetylation plays a critical role in EKLF's ability to accessits target site within chromatin in vivo, unlike the in vitrodata with naked DNA. Since EKLF is acetylated at this residueby CBP, these data also imply that in vivo, DNA-bound EKLF mustbe highly enriched for its acetylated form.

View larger version (34K):
[in this window]
[in a new window]

FIG. 3. EKLF expression leads to coactivator recruitment and chromatin remodeling at the β-globin locus. (A) ChIP of EKLF binding to the β-major promoter was monitored in 18SCF cell lines not infected (–) or infected with wild-type or acetylation mutant EKLF as indicated. Occupancy by wild-type EKLF and the 302K/R mutant is observed at the adult globin promoter. The occupancy levels of 288K/R and 288/302/K/R and the 18SCF lines are similar to that of the control without antibody (No Ab). Error bars show standard deviations. (B) Recruitment of CBP to the adult β-globin promoter and to two of the HSs (HS2 and HS3) at the LCR (indicated by asterisks in the schematic diagram on top) was monitored in 18SCF cell lines not infected or infected with wild-type or mutant EKLF as indicated. CBP is enriched at the globin promoter, HS2, and HS3 in the cell lines expressing wild-type EKLF and the 302K/R mutant. The levels of CBP at HS2, HS3, and the globin promoter in the 288K/R and 288/302/K/R mutant are similar to the level in the parent cell line. Error bars show standard deviations. Ab, antibody; –, absent (Ab) or not infected (Rescue); maj, major; min, minor. (C) A DNase hypersensitivity assay was performed to determine the extent of chromatin remodeling at the HS3 site of the LCR. Nuclei were extracted from 18SCF cells not infected (–) or expressing wild-type or mutant EKLF (as indicated) and subjected to digestion with increasing amounts of DNase I (indicated in units [U]) for a constant time, followed by DNA extraction and quantitative PCR with equal amounts of DNA. The amount of amplicon remaining in each case was normalized to the amount of DNA in undigested starting material (UD), which was set to a level of 100%. The neurofilament subunit gene (Nf-M) served as a negative control. The data are representative of the results of two experiments. WT, wild type.

The EKLF-CBP interaction is crucial for EKLF's optimal functionas an activator (77), and CBP is not present at the β-globinpromoter in EKLF-null cells (7). EKLF binds to CACCC sequencesin HS2, HS3, and the β-major promoter (28). With this inmind, we tested the recruitment of CBP to the β-globinregion by using ChIP after EKLF rescue of 18SCF cells. We foundthat CBP is enriched at the β-major promoter, HS2, andHS3 regions only in lines expressing wild-type EKLF and thelysine 302 mutant, but not in the lines expressing lysine 288and the double mutant (Fig. 3B).

Considering that the SWI/SNF complex has been demonstrated tomediate chromatin remodeling by EKLF in vitro and that thisis dependent on its acetylation status (4, 34, 77), we nexttested whether the chromatin structure was affected in the rescuedcells by performing an in vivo DNase I hypersensitivity assay.We focused on the status of chromatin remodeling at HS3 as indicativeof the status of an active chromatin hub (19) and also becausethis site is lost in the absence of EKLF (71, 74). Cells expressingwild-type and mutant EKLF were harvested, and the nuclei weredigested with increasing amounts of DNase I. The DNA was subsequentlypurified, and the presence of HS3 was analyzed by using real-timePCR. The parental 18SCF cell line and the promoter of anothergene (encoding Nf-M, a neurofilament subunit) not activatedby EKLF served as negative controls. The results of these analysesdemonstrate that rescue with wild-type EKLF leads to HS formationwith HS3, as the rescued line is $~$ 5- to 10-fold more sensitiveto DNase I (i.e., fewer PCR amplicons remain after treatment,indicating an open chromatin structure) (Fig. 3C). This is alsoobserved after rescue with the EKLF 302K/R mutant (Fig. 3C).However, rescue with EKLF 288K/R is virtually indistinguishablefrom the parental line (Fig. 3C). The results of these studiesdemonstrate that the EKLF acetylation status, particularly atan available lysine 288 residue, is of primary importance forbinding to DNA and for contributing to the ultimate formationof an open chromatin structure in vivo.

Altered acetylation of histone H3 at the β-globin promoter is dependent on EKLF. A major factor regulating the degree of chromatin folding ishistone acetylation (65). As we found that there was less recruitmentof CBP to the β-globin locus and less of an open chromatinstructure at HS3 in the cells expressing 288K/R mutant EKLF,we also tested the status of histone acetylation at the β-majorpromoter and the LCR in two ways. First, we performed ChIP usingpan-acetyl-H3 antibodies on the rescued 18SCF cell lines expressingwild-type or 288K/R, 302K/R, or 288/302/K/R mutant EKLF. Wefound that the β-globin promoter, as well the HS2 and HS3sites, had a large enrichment of acetyl-H3 in cell lines expressingwild-type and 302K/R EKLF (Fig. 4A). However, the lines expressing288K/R and 288/302/K/R were similar to the negative controlsand exhibited little, if any, acetyl-H3 enrichment (Fig. 4A).

View larger version (21K):
[in this window]
[in a new window]

FIG. 4. EKLF-dependent histone H3 acetylation at the β-globin locus. (A) ChIP was performed to determine the status of histone H3 acetylation at the β-globin locus in 18SCF EKLF-null cell lines before and after rescue with wild-type or acetylation mutants of EKLF. ChIPs were performed in the absence or in the presence of pan-acetyl-H3 antibody (acH3) (product no. 17-615; Millipore). Histone modification associated with HS2 and HS3 of the LCR and the β-promoter (β-major) was determined by quantitative real-time PCR. –, absent (Ab) or not infected (Rescue); WT, wild type. (B) MEL cells containing a stably integrated shEKLF construct were treated with 2 µg/ml Dox for 48 h to induce its expression (which results in knockdown of EKLF levels to $~$ 20% [8]), followed by treatment with 1.5% DMSO to induce terminal differentiation. After 96 h, ChIP was performed either without antibody or in the presence of pan-acetyl-H3 antibody (acH3) (product no. 17-615; Millipore) or acetyl-histone H3 Lys14 antibody [acH3(lys14)] (product no. 07-353; Millipore), and modifications associated with HS2 and HS3 of the LCR and the β-promoter (β-maj) were determined by quantitative real-time PCR. The results from three independent experiments have been averaged. Error bars show standard deviations. –dox, without Dox; +dox, with Dox; IgG, immunoglobulin G; Ab, antibody.

Second, we used an MEL cell line that contains a doxycyclin(Dox)-inducible EKLF-specific shRNA (8) to monitor effects beforeand after loss of EKLF. MEL cells normally express high levelsof EKLF message and protein even in the absence of dimethylsulfoxide (DMSO) or hexamethylene bisacetamide induction, afterwhich β-globin expression and terminal differentiationensues (49). Since CBP acetylates histone H3 at lysine 14 (37),a prediction from our data is that loss of EKLF should resultin a decrease in the level of histone H3 K14 acetylation dueto the absence of CBP recruitment by EKLF. The stably modifiedMEL cells were grown for 2 days in the presence of 2 µg/mlDox to induce shEKLF. This treatment is sufficient to decreaseEKLF message and protein and β-globin message levels byover 80% (8). Terminal differentiation was then induced withDMSO to establish active transcription of the β-globingene. After 4 days of differentiation, cells were subjectedto ChIP using pan-acetyl-H3 and acetyl-H3 Lys14-specific antibodies.Consistent with the results of our rescue studies, we foundthat the decrease of EKLF protein levels led to a major lossof acetylated H3 protein at the promoter and HS2 and HS3 (Fig.4B). The histone acetylation status on the necdin gene remainedunaltered, showing that this is not just a general effect. However,we found an even more profound loss of acetyl-H3 lys14 acrossall sites examined (Fig. 4B).

Together, these two sets of data with different cell lines demonstratethat the EKLF protein is critical for the establishment of theproper level of acetylated H3 at different regions of the β-globinlocus and further establish that its interaction with and recruitmentof CBP are critical for the downstream modification of lysine14 of histone H3, thus confirming our hypothesis that EKLF recruitsCBP to bring about modifications of the chromatin that favortranscription.

The EKLF zinc finger domain interacts with the amino-terminal tail of core histone H3. Our current results and previous observations indicate thatacetylated EKLF influences chromatin structure and gene expressionby recruiting chromatin modifiers, such as CBP, and remodelers,such as the SWI/SNF complex (4, 34). In light of this, we canhypothesize that the core histones might provide a nucleosomaltarget for chromatin opening by EKLF. The interaction of otheractivators with nucleosomal components (3, 50) led us to testthe interaction of EKLF with the core histone proteins (H2A,H2B, H3, and H4) by performing an in vitro pull-down assay usingbacterially expressed, GST-tagged EKLF and individually in vitro-translated,[³⁵S]methionine-labeled core histone proteins. We find thatEKLF preferentially interacts with core histone H3 (Fig. 5A).

View larger version (31K):
[in this window]
[in a new window]

FIG. 5. Interaction between EKLF and histone H3. (A) A pull-down assay was performed using equal amounts of GST or GST-EKLF (not shown) coupled to glutathione Sepharose beads by incubation with in vitro-translated [³⁵S]methionine-labeled histones H2A, H2B, H3, and H4. Ten percent of the labeled histone input protein (In) was also included in the gel analysis. (B) A pull-down assay using equal amounts of GST, GST-EKLF, GST-proline domain (Pro) or GST-zinc finger domain (ZnF) (as determined by Coomassie staining after SDS-PAGE [bottom]) was performed with in vitro-translated [³⁵S]methionine-labeled H3.1 and H3.3 as indicated on the right. Ten percent of the labeled histone input protein (In) was also included in the gel analysis. (C) A pull-down assay using equal amounts of GST, GST-H3.1 N-terminal domain (amino acids [aa] 1 to 41), GST-H3.3 N-terminal domain (aa 1 to 41), or GST-H3.3 globular domain (aa 42 to 136) (as determined by Coomassie staining after SDS-PAGE [bottom]; GST material is shown in panel B, which is from the same gel) was performed with in vitro-translated [³⁵S]methionine-labeled full-length EKLF. Ten percent of the labeled EKLF input protein (In) was also included in the gel analysis. All samples (in panels A to C) were analyzed by SDS-PAGE, and the gels were treated with EN³HANCE solution (NEN Life Science) prior to detection of radioactivity by autoradiography. (D) In vivo coimmunoprecipitation of EKLF and histone H3. Full-length (wt), MZn1-3 (mut ZnF 1,2,3), 288K/R, and 302K/R variants of EKLF were cotransfected into 293T cells along with either FLAG-HA-H3.1 or FLAG-HA-H3.3. After immunoprecipitation of extracts with M2-agarose (Flag) or immunoglobulin G (IgG) (indicated on left), samples were resolved on SDS-PAGE gel and Western blots were probed with either EKLF polyclonal or anti-HA antibodies (indicated on right). Ten percent of total protein extract served as Input. Fl, FLAG; +, present; –, absent; mut ZnF, mutant zinc finger.

Four different histone H3 variants have been reported (H3.1,H3.2, H3.3, and CENPA) (63, 64). A large number of recent studieshave focused on the role of variant histones in marking transcriptionalstates. The histone variant H3.3 is incorporated into nucleosomesindependent of DNA replication and is a potential mediator ofepigenetic memory of the active transcriptional state (1, 13,47). However, the deposition of H3.1 is tightly coupled to DNAreplication (1). Even though H3.1 and H3.3 differ by only fiveamino acids (41, 63), four of these residues reside in theirglobular domain and could be crucial for their distinctive depositionin the chromatin and for their interaction with protein partners(69). We had used the histone variant H3.3 in our initial experimentto investigate whether EKLF interacts with any of the core histoneproteins. To see if EKLF also interacts with the core histoneH3.1 and, also, to map the interaction domain of EKLF, we performedpull-down assays using either GST-tagged full-length EKLF protein,the GST-tagged EKLF proline domain, or the GST-tagged EKLF zincfinger domain, along with labeled, in vitro-generated H3.1 orH3.3 protein. We found that EKLF interacts equally well withH3.1 and H3.3 and that the interaction is mediated via the zincfinger domain of EKLF (Fig. 5B).

We next mapped the region of histone H3 that is responsiblefor its interaction with EKLF. In this case, we used the GST-taggedhistone H3.3 tail or its globular domains and full-length EKLFand found interaction only with the amino-terminal tail (Fig.5C). Not surprisingly, a similar level of interaction is seenwith the amino-terminal tail of histone H3.1. An analogous experimentperformed between the zinc finger domain of EKLF and N-terminalor globular histone H3 domains also leads to the same conclusion(data not shown). Although at this level of resolution we seeno evidence of discrimination between the two histones, we canconclude that EKLF specifically interacts with the amino-terminalregion of histone H3 via its zinc finger domain.

The results of our studies show that EKLF interacts with CBPand that lysine 288 is necessary and important for this interaction.We have also shown that EKLF interacts with the N-terminal tailof histone H3 through its zinc finger domain. This promptedus to determine whether the EKLF acetylation mutant that iscompromised in its ability to recruit CBP is also affected inits ability to interact with the histone. Additionally, sincethe zinc finger region of EKLF is also its DNA-binding domain,we wanted to determine whether the ability of EKLF to bind toDNA and interact with histone H3 can be uncoupled. For thiswe used a mutant EKLF that is compromised in its ability tobind to DNA (12). Wild-type EKLF or 288K/R, 302K/R, or MZn1-3(containing mutations that alter critical histidines known tobe important for coordinating zinc within each finger) variantswere cotransfected with either FLAG-hemagglutinin (HA) H3.3or FLAG-HA H3.1 in 293T cells. After coimmunoprecipitation usingM2-agarose or immunoglobulin G and Western blot analysis usinganti-EKLF antibody, we found that the wild type and mutantsof EKLF all attain a similar in vivo level of interaction withhistone H3.3 or H3.1 (Fig. 5D). This enables us to concludethat neither the acetylation status nor the zinc finger structureof EKLF plays a role in its ability to interface with histoneH3.

Distribution of histone H3.3 in mouse β-globin locus in MEL cells. It has been reported from studies with Drosophila melanogasterthat although no specific posttranslational modification isrestricted to either H3 variant, H3.3 is enriched in modificationsassociated with transcribed genes, including acetylation oflysine 14. In contrast, protein modifications associated withrepressed genes are present on H3.1 (47). In mammalian celllines, H3.3 deposition is replication independent and occurson actively transcribed genes, but not on silent loci (13, 16).In a chicken hematopoietic cell lineage, however, there is nostraightforward correlation between sites of H3.3 incorporationand active or repressive chromatin modifications (30). To investigatethis issue within the mammalian hematopoietic lineage, we constructedMEL cell lines that stably express either FLAG-HA-H3.1 or FLAG-HA-H3.3(51, 69). We screened individual colonies and pools for linesthat expressed equal levels of the tagged protein (Fig. 6A)to rule out the possibility that increased or decreased levelsof expression were responsible for any differential incorporationof one histone versus the other. We then used these lines totest the in vivo interaction of histone H3 variants with EKLFvia coimmunoprecipitation. Similar to our in vitro results,we found that EKLF interacts at a comparable level with eachH3 variant and that this level of interaction does not changeafter the induction of terminal differentiation and β-globingene expression (Fig. 6A). Finally, to test the relative distributionof H3.1 or H3.3 across the globin locus, we performed ChIP withMEL cells stably expressing H3.3 or H3.1 before and after treatmentwith DMSO to cause differentiation and the expression of adultβ-globin. After 4 days, cells were collected and ChIP wasperformed using M2-agarose (FLAG antibody) to monitor the presenceof the tagged H3 variant and pan-acetyl-H3 antibody to monitorthe total (tagged and untagged) acetylated histone H3 detectedunder each condition within each cell line. We monitored occupancyat the embryonic βH1 and ${varepsilon}$ -globin (Ey) promoters and theadult β-major promoter for enrichment of either variantof histone H3. The necdin promoter served as a control as itis nonerythroid and should not be affected by the onset of erythroiddifferentiation. Under conditions in which pan-acetyl-H3 isenriched in both cell lines, we found that only H3.3 is enrichedon the adult β-major promoter after the differentiationof MEL cells, but not on the embryonic βH1 or ${varepsilon}$ -globin promoters(Fig. 6B). This suggests that, similar to the results of studiesshowing a differential utilization of histone H3 variants, histoneH3.3 is specifically enriched during active transcription ofthe β-like globin locus in the erythroid lineage.

View larger version (20K):
[in this window]
[in a new window]

FIG. 6. Distribution of H3.1 or H3.3 across the β-like globin locus in MEL cells. (A) 745A MEL cell lines stably transfected with FLAG-HA-histone H3.1 or FLAG-HA-histone H3.3 were subjected to immunoprecipitation with M2-agarose (Sigma) before (–) or after (+) induction of differentiation with DMSO for 96 h. Samples were resolved by SDS-PAGE, followed by immunoblotting using anti-HA antibody, anti-pan-acetyl histone H3 antibody (product no. 17-615; Millipore), or anti EKLF polyclonal antibody as indicated. (B) MEL cells containing stably expressed FLAG-HA-histone H3.1 or FLAG-HA-histone H3.3 were analyzed before (– differentiation) or after (+ differentiation) DMSO treatment by performing ChIP with anti-FLAG M2-agarose (Flag) or pan-acetyl H3 antibody (acH3) (product no. 17-615; Millipore) or without antibody (–) by real-time PCR using primers specific for adult β-globin (β-maj), embryonic βH1, embryonic ${varepsilon}$ -globin (Ey), and necdin promoters (see Fig. 3B). All experiments were performed in triplicate. Error bars show standard deviations. Fl, FLAG; Ab, antibody; maj, major.

DISCUSSION

Top

DISCUSSION

Sequence-specific DNA-binding transcription factors utilizevarious molecular mechanisms to modulate chromatin architectureand recruit the basal transcription machinery (35, 39, 52).Here we propose a mechanism by which a lineage-restricted transcriptionfactor regulates the formation of tissue-specific chromatindomains. For this study, we used an EKLF^–/– fetalliver cell line, in combination with retroviral rescue of itsbiologically relevant target, to explore how the acetylationof EKLF contributes to its chromatin remodeling ability at theadult β-globin locus in vivo. By focusing on effects atthe endogenous locus, this system avoids the deficiencies ofusing a nonchromatinized reporter assay with overexpressed transfectedconstructs. Given EKLF's critical role in erythroid gene regulation,this is a particularly good model system in which to study chromatinremodeling and the transcriptional competence of the β-globinlocus.

Acetylation of EKLF is essential for transcription and chromatin remodeling. Our laboratory has previously demonstrated that EKLF binds toCBP (76, 77), a cofactor whose acetyltransferase activity resultsin the acetylation of EKLF at two specific sites, lysine 288and lysine 302. CBP also acetylates histone H3 (37). We findthat these effects are interrelated, as the CBP-dependent acetylationof EKLF at 288K is itself necessary for the recruitment of CBPto the β-globin promoter, an association that remains criticalfor its subsequent acetylation of histone H3 at lysine 14. Consistentwith this is the fact that in the absence of EKLF, the acetylationof histone H3 at lysine 14 is more profoundly affected. Thissuggests that EKLF must interact with and be modified by CBPprior to its binding to DNA and that the ability of CBP to modifyboth transcription factor and histone substrates is equallynecessary for recruitment and their proper arrangements at thelocus. The acetylation of EKLF lysine 288 also provides a more-efficientsubstrate for EKLF association with the SWI/SNF chromatin remodelingcomplex (4, 77), likely by means of its interaction with theBRG1 component (10, 33, 34). Coupled with the observation thathistone H3 acetylation at lysine 14 aids in the recruitmentof TFIID (73), the result is the formation of a transcriptionallycompetent, three-dimensional epigenetic protein/DNA complexthat has been called the "active chromatin hub" (18, 19).

DNA binding by EKLF in vivo is affected by acetylation. The acetylation of transcription factors can affect their DNA-bindingability, nuclear localization, protein stability, or protein-proteininteractions (36, 38). Abolishment of the acetylation sitesin EKLF has no effect on nuclear localization or protein stability(61, 62), and the acetylation status of EKLF does not affectits interaction with DNA in vitro (77). Yet, unexpectedly, wenow show that mutation at the critical 288K acetylation sitealters EKLF's interaction with chromatin in vivo. This suggeststhat the mode of binding by EKLF to naked DNA in vitro is differentfrom its mode of binding to chromatinized templates in vivo.Similar results have been obtained with GATA-1, where lack ofacetylation leads to differential binding with chromatinizedor naked templates (38). It is relevant to note that other mutationsin GATA-1 that do not affect its binding to DNA also show diminishedchromatin occupancy at some sites in vivo (32).

A model for sequential epigenetic modification and transcriptional activation. The carboxy-terminal DNA-binding domain (amino acid residues293 to 376) of EKLF is sufficient to induce complete chromatinremodeling of a β-promoter template in vitro (4). We haveshown that this region of EKLF also binds to histone H3. Furthermore,this interaction does not require an intact zinc finger structure,suggesting that the binding of EKLF to histone and DNA can occurindependently of each other. The chromatin of the β-likeglobin cluster is open well before the onset of transcription(6, 29). Together, these observations raise the possibilitythat EKLF might be playing a DNA-independent role at this earlystage and compels us to consider a model that is surprisingbut consistent with our observations. Our hypothesis is thatEKLF interacts with CBP, is acetylated by it, and recruits itto the globin promoter. EKLF directly interacts with histoneH3 and thus serves as a bridge or adapter between CBP and histoneH3, facilitating the interaction between the two. This in consequenceleads to active histone modifications (acetylation) and subsequentchromatin remodeling at the locus, all of which occurs in theabsence of direct DNA binding by EKLF. Only after the remodelinghas occurred can EKLF fully access and bind to its cognate DNA(CACCC) element at the β-major promoter, thus further stabilizingthe active chromatin structure (hub) and ultimately leadingto gene transcription. This process cannot begin in the absenceof CBP acetylation of EKLF 288K and cannot finish without directrecognition of the DNA by EKLF. Alternatively, acetylated 288Kmay itself be required for the stable association of EKLF bindingto chromatinized DNA at the promoter, with this initial stableassociation being necessary for subsequent remodeling and transcription(42, 43). Since the amino terminus of the histone proteins isthe main site for posttranslational modifications, the interactionof EKLF with the amino-terminal tail of histone H3 supportsour idea that physical interaction between EKLF and the H3 tailis required for its acetylation and is a prerequisite for genetranscription. This model also raises the possibility that EKLF/CBPis part of a larger EKLF-containing protein complex that maycontain (an)other DNA-binding protein(s). Ordered recruitmentcan be accomplished in multiple ways depending on the gene thatis to be regulated (15), and the dynamic nature of histone modificationsand ATP-dependent nucleosomal sliding at the promoter (48) lendcredence to these ideas. Our model is also broadly consistentwith the results of recent GATA-1 rescue studies (reviewed inreference 75).

Other proteins that bind the EKLF zinc fingers do not necessarilyrequire an intact structure for this interaction. For example,Sin3A/histone deacetylase 1 binds EKLF and aids its repressionactivity in the absence of an EKLF finger structure (12), andEKLF nuclear import is (mostly) retained in the absence of anEKLF finger structure that constitutes its major nuclear localizationsignal (58, 62).

Genetic manipulation of CBP and BRG1 demonstrate that they areessential for erythroid development (5, 11, 25, 55). Their requirementcan be explained (at least in part) by the results of our presentstudies, where we have found at a molecular level that CBP isresponsible for EKLF and histone H3 acetylation, steps thatwe suggest are needed for the ultimate chromatin remodelingcatalyzed by SWI/SNF that occurs at the β-like globin locus.At the same time, the recruitment of CBP to the locus will notoccur in the presence of EKLF that cannot be itself acetylatedby CBP at 288K. As a result, CBP's ability to transfer acetylgroups both to histones and to transcription factor substratesis critically important for successful gene activation. Suchcovalent modifications contribute significantly to the overallmechanism of transcriptional activation at specific genes andhelp explain the vital role of CBP in this process.

The binding of EKLF to histone H3 led us to examine the interactionof EKLF with the two best-characterized H3 variants. Since H3.3is the histone protein associated with actively transcribingDNA, it seemed reasonable that EKLF should preferentially interactwith it. But under the conditions tested, either in vitro, intransient transfection, or in stable lines, we did not findthis to be true. However, using tagged H3.3 and H3.1 stablelines to monitor whether either of the H3 proteins was preferentiallyrecruited to the ${varepsilon}$ -globin, β-H1, and necdin promoters (alltranscriptionally inactive in differentiated MEL cells) andthe β-major promoter (transcriptionally active in differentiatedMEL cells), we found enrichment of H3.3, but not H3.1, at theactively transcribing gene, similar to that seen at other promoters(13) and consistent with the idea that the intrinsically lessstable H3.3-containing nucleosome is preferentially incorporatedat gene-transcribing regions (31). The further study of theeffects of this and other histone variants on β-like globingene expression and whether EKLF plays any role in this differentialrecruitment will allow us to ultimately distinguish variousmechanisms of promoter activation that may be operant acrossthe locus and that lead to its precise developmental regulation.

ACKNOWLEDGMENTS

Wenjun Zhang and Felix Lohmann were involved in the initialphase of this study. We thank members of the Bieker lab fortheir ongoing support and discussion, and especially Felix Lohmannfor his help with the ChIP assays and Saghi Ghaffari and JohnCrispino for help with the retroviral infections. We thank K.Wilson, Y. Nakatani, and N. Mermod for plasmid constructs, F.Morle for the shEKLF MEL cell lines, and Lee Wall and his labfor help in establishing the 18SCF cell line.

This work was supported by NIH PHS grant R01 DK46865 to J.J.B.The Quantitative PCR Shared Research Facility is supported byMount Sinai School of Medicine.

FOOTNOTES

* Corresponding author. Mailing address: Mount Sinai School of Medicine, Department of Developmental and Regenerative Biology, Box 1020, One Gustave L. Levy Place, New York, NY 10029. Phone: (212) 241-4143. Fax: (212) 860-9279. E-mail: james.bieker{at}mssm.edu

Published ahead of print on 18 August 2008.

REFERENCES

Top