| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Published online before print March 8, 2007
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
and C/EBP
Cedars-Sinai Medical Center, Division of Hematology/Oncology, Burns and Allen Research Institute and David Geffen School of Medicine at University of California Los Angeles, Los Angeles, California, USA
1 Correspondence: Cedars-Sinai Medical Center, Division of Hematology/Oncology, Davis Bldg. 5019, 8700 Beverly Blvd., Los Angeles, CA 90048, USA. E-mail: gombarta{at}csmc.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. Dimerization required the ZIP motif of each protein and redirected DNA binding of C/EBP and ATF4 from their respective symmetric consensus sites to asymmetric C/EBP and cAMP response element sites. The C/EBP:ATF4 heterodimer bound to the C/EBP sites in the promoters of the myeloid-specific genes encoding neutrophil elastase (NE) and the G-CSF receptor (G-CSFR). Also, the heterodimer bound a previously uncharacterized site in the promoter of the mim-1 gene at nucleotide –174. Coexpression of ATF4 and C/EBP in the presence of c-Myb synergistically activated the mim-1 and NE promoters compared with C/EBP plus c-Myb alone. Synergistic activation was not observed for the G-CSFR promoter and only occurred in the presence of c-myb with the NE or mim-1 promoters. In contrast, ATF4:C/EBP
dimers bound to the C/EBP sites in the G-CSFR and NE promoters, but transcriptional activation was inhibited by 30–80% in the presence or absence of c-Myb. We propose that ATF4 may regulate myeloid gene expression differentially by potentiating C/EBP but inhibiting C/EBP-mediated transcriptional activation.
Key Words: CCAAT/enhancer • yeast two-hybrid • activating transcription factor • heterodimer
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, C/EBPß [nuclear factor-IL-6, C-reactive protein 2 (CRP2)], C/EBP
(nuclear factor-IL-6ß, CRP3), C/EBP
, C/EBP (CRP1), and growth-arrest and DNA-damage induced 153 (GADD153)/C/EBP homologous protein-10 (CHOP 10) [5
]. Most members are expressed in a wide variety of tissues and are critical for processes ranging from signal transduction to adipocyte, hepatocyte, and hematopoietic cell differentiation [6
7
8
9
10
]. In the hematopoietic system, the C/EBP proteins are expressed primarily in monocytes, macrophages, granulocytes, and their precursors, indicating that they play an important role in myeloid development [8
, 11
, 12
]. C/EBP and - levels increase during granulocytic but not monocytic differentiation [8
, 13
14
15
]. Myeloblastic and promyelocytic cell lines representing immature cells of the granulocytic lineage display the highest levels of C/EBP and - [8
, 14
, 15
]. Mice lacking the C/EBP and - proteins displayed defects in granulopoiesis and failed to generate functional neutrophils and eosinophils [10
, 16
]. Mutations in C/EBP are responsible for the development of neutrophil-specific granule deficiency [17
, 18
]. Mutations in C/EBP are involved in the development of acute myeloid leukemia (AML) [19
, 20
].
Unlike other family members, expression of C/EBP is restricted to myeloid lineage cells [14
, 15
, 21
, 22
]. There are four C/EBP protein isoforms of calculated MW 32.2, 30.0, 27.8, and 14.3 kDa [15
, 22
]. The 14.3-kDa isoform completely lacks a TAD [15
]. The 32- and 30-kDa isoforms activate transcription of important neutrophil and eosinophil granule genes and promote granulocytic differentiation [23
, 24
], and the 27-kDa form represses eosinophil major basic protein gene expression [25
].
A number of myeloid-specific genes contain functional C/EBP-binding sites in their promoters. These include the primary granule proteins neutrophil elastase (NE), proteinase 3, and myeloperoxidase (MPO), the receptors for G-CSF (G-CSFR), M-CSFR, and GM-CSFR, and the secondary granule protein lactoferrin [26
27
28
29
30
]. Also included are the avian genes encoding the mim-1 and myeloid growth factor (MGF) proteins [7
, 31
]. The C/EBP protein cooperates with PU.1 to activate the G-CSFR and GM-CSFR genes and with RUNX-1 to activate the M-CSFR gene [26
, 29
, 30
]. The C/EBP protein activated transcription from the promoter of the G-CSFR gene but less effectively than C/EBP and increased transcriptional activity from the human MPO promoter when cotransfected into myeloid cell lines [15
, 22
]. Together with c-Myb, C/EBP and - cooperatively activated the NE promoter [32
, 33
]. The cooperative transcriptional activation by the C/EBP proteins and c- or v-Myb was first described for the avian myeloid-specific mim-1 promoter. When coexpressed with Myb, the avian homologue of human C/EBPß, middle weight neurofilament (NF-M), induced expression of the myeloid-specific genes mim-1, chicken MGF, and lysozyme in myeloid and nonmyeloid cells [7
, 34
, 35
]. Also, C/EBP cooperatively activated transcription from the mim-1 gene promoter together with c-Myb [33
].
The C/EBP and - proteins show overlapping expression during myeloid-lineage development and are especially critical for normal granulopoiesis [5
, 10
, 14
, 15
, 36
]. As both proteins bind the same C/EBP sites, they potentially regulate the same genes. The ability of bZIP family members to heterodimerize could allow a possible third partner to regulate the function of C/EBP or -. In this report, we demonstrate that the activating transcription factor 4 (ATF4) protein heterodimerizes with C/EBP and - [37
38
39
]. This interaction transforms C/EBP into a potent transcriptional activator and represses C/EBP-mediated transcription from promoters of several myeloid-specific genes. Our results implicate ATF4 as a potential regulator of C/EBP-mediated, myeloid gene transcription.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Yeast two-hybrid screen
The entire coding region (amino acid residues 1–249), the amino-terminal half (residues 1–115), and the bZIP domain (amino acid residues 147–249) of C/EBP were amplified by PCR and subcloned into pGBT8 to generate a fusion protein with amino acid residues 1–147 of the GAL4 DBD (GAL4DBD; see Fig. 1A
).
|
30 or pGBT8-C/EBP147–249 as described (Clontech). Cotransformants containing interacting fusion proteins were selected on SD medium lacking Trp, Leu, and His and containing 5 mM 3-amino trizol. The ß-galactosidase-positive clones were isolated, and the inserts were amplified by PCR using Matchmaker 5' and 3' AD LD-insert screening amplimers (Clontech). The PCR products were sequenced and analyzed by BLAST searches against nucleotide and protein databases [40
, 41
]. The ß-galactosidase filter and liquid assays were used to measure the strength of interactions and were performed as described by the manufacturer (Clontech).
GST and maltose-binding protein (MBP) fusion protein pull-down assays
The GST fusion proteins GST-C/EBP1–115 and GST-C/EBP147–249 (see Fig. 1A
), GST-ATF425–351 (see Fig. 1B
), and GST-C/EBP (rat) full-length fusion [6
] were generated using the pGEX vectors (Pharmacia, Uppsala, Sweden). GST fusion proteins with full-length CREB1 and cAMP-responsive element (CRE)-BP-1 (ATF2) were kindly provided by Richard Gaynor (Eli Lilly, Indianapolis, ID, USA) [42
]. The full-length C/EBP PCR product was cloned into pMALc2 (New England Biolabs, Inc., Beverly, MA, USA) to generate a MBP-C/EBP30 fusion. The GST fusion proteins were expressed and purified using glutathione-sepharose as described (Pharmacia). The MBP fusion was expressed and purified using amylose-resin as described by the manufacturer (New England Biolabs, Inc.).
For pull-down assays, plasmids encoding the proteins were translated in vitro using the transcription-translation-reticulocyte lysate system (Promega Biotech, Madison, WI, USA) in the presence of 35S-methionine (Dupont NEN, Wilmington, DE, USA) as described by the manufacturer. The plasmids included pcDNAI-C/EBP30 [22
], pcDM7-CREB2/ATF4, and pcDM7-CREB2/ATF4249–351 [38
], the latter two kindly provided by Jeffrey Leiden (Abbott Laboratories, Abbott Park, IL, USA), and pcDNAI-C/EBP (murine) [43
], generously provided by Kleanthis Xanthopolous (Anadys Pharmaceuticals Inc., San Diego, CA, USA). The radiolabeled proteins were mixed with glutathione-sepharose or amylose-resin loaded with GST- or MBP-fusion proteins and processed as described previously [44
].
EMSAs
For EMSAs, double-stranded oligonucleotide probes (10 pmole) were end-labeled using T4 polynucleotide kinase and [32P]ATP as described by the manufacturer (Invitrogen). Probes (0.1 pmole/reaction) were mixed with purified fusion proteins. The total amount of protein in the reaction was adjusted to 200 ng with purified GST protein. Polyinosinic-polycytidylic (Pharmacia) and BSA (Sigma Chemical Corp., St. Louis, MO, USA) were added to 50 µg/ml and 300 µg/ml final concentrations, respectively. Binding reaction conditions were 20% glycerol, 20 mM HEPES (pH 7.9), 50 mM NaCl, 2 mM MgCl2, and 1 mM DTT. Competition experiments were performed using oligonucleotides representing an unlabeled wild-type or nonspecific, double-stranded oligonucleotide at a 100:1 molar ratio. Reactions were incubated at room temperature for 30 min and analyzed by gel electrophoresis through a 4% polyacrylamide gel using a high ionic-strength Tris-glycine buffer (50 mM Tris, 400 mM glycine, 1 mM EDTA, with pH adjusted to 
8.5). Gels were exposed to Kodak X0-Mat film.
Double-stranded oligonucleotides for CRE sites from the promoters of the somatostatin (SOM), collagen gene 8 (COL-8), phosphoenolpyruvate carboxykinase (PEPCK), and the enkephalin (ENK) genes were described previously [45 ]. Oligonucleotides for C/EBP sites from the NE and G-CSFR gene were described previously [33 ]. The oligonucleotides for mim-1 were Mim-60 (5'-ACTGATTGGCCAACACAACAG-3'), Mim-160 (5'-CCTGTCTTTCCCAACCAGCTC-3'), and Mim-174 (5'-AAGACACCCGTTACTTTACCTGTC-3') [46 ]. The C/EBP consensus (5'-TGCAGATTGCGCAATCTGCA-3') was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA) [29 , 32 ]. The chimeric site was described previously [47 ]. The nonspecific competitior (5'-TCGAGACGTCTTTGACTCGCTCAAAG-3') was derived from a site in the mim-1 promoter, which C/EBP or ATF proteins were unable to bind.
For supershift analysis, oligonucleotide and protein were incubated for 15 min at room temperature, then antibodies were added, and the reactions were incubated for an additional 15 min before electrophoresis. Rabbit polyclonal antibodies against CRP1 (rat C/EBP) and human CREB2 (ATF4) were purchased from Santa Cruz Biotechnology.
Transfections, coimmuoprecipitation analysis, and reporter assays
Approximately 3 x 106 COS-1 cells were electroporated with 15 µg DNA in 0.4 cm cuvette with 0.7 ml DMEM containing 10% FBS (1.5 kV, two 90-msec pulses) using the T820 Electroporator (Genetronics BTX, San Diego, CA, USA). Cells were plated in 10 ml complete medium, incubated 48 h, and harvested for immunoprecipitation and Western blot analysis as described previously [48
]. Antibodies against ATF4 are described above, and anti-C/EBP was described previously [22
]. For immunoprecipitation of heterodimer complexes, 20 pmole double-stranded chimeric oligonucleotide and 1 µg antibody were added to 150 µg cell lysate. Immunoprecipitated complexes were subjected to Western blot analysis.
Promoter-reporter assays in Jurkat cells were performed as described previously [49 ]. The promoter-reporter constructs pMim-Luc (and mutants), pNE-Luc, and (–74 to +67) G-CSFR-pXP2 were kind gifts from Achim Leutz (Max Delbruck-Centrum fur Molekulare Medizin, Berlin, Germany), Alan Friedman (Johns Hopkins University, Baltimore, MD, USA), and Daniel Tenen (Harvard Medical School, Boston, MA, USA), respectively.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
interaction partners147–249 fusion, indicating that the ZIP motif of C/EBP was involved in the interaction (Fig. 1A)
. The shortest insert for ATF4, which was isolated, encoded amino acids 110–351 (Fig. 1B)
. All isolated clones contained the bZIP domain of ATF4, but the TAD was not required. Liquid ß-galactosidase assays demonstrated that the interaction between full-length C/EBP and ATF4 was 40% stronger than the interaction between C/EBP147–249 and ATF4 (Table 1
). This suggested that amino acid residues N-terminal to the bZIP domain of C/EBP enhanced heterodimerization.
|
induction (Fig. 1D)
. Levels of ATF4 mRNA did not change. Induction of monocytic differentiation in HL-60 with 1,25(OH)2D3 did not alter ATF4 levels (Fig. 1D)
. The levels of ATF4 protein did not change in response to ATRA or 1,25(OH)2D3 (Fig. 1E
, right panel).
ATF4 and C/EBP dimerization requires the bZIP domain
To verify and characterize the ATF4:C/EBP interaction, we performed in vitro pulldown assays (Fig. 2A
). The MBP-C/EBP30 fusion but not MBP alone pulled down the in vitro-synthesized, 35S-methionine-labeled, full-length and truncated ATF4249–351 (Fig. 2A
, Lanes 6 vs. 5 and Lanes 8 vs. 7, respectively). The full-length MBP-C/EBP fusion protein specifically pulled down the 35S-labeled C/EBP30 (Fig. 2A
, Lanes 4 vs. 3). No products were pulled down in the reticulocyte lysates programmed with the empty vector (Fig. 2A
, Lanes 1 and 2). The immobilized MBP- fusion protein retained 30% of the input C/EBP or ATF4 protein. This suggested that the affinity of C/EBP for itself or ATF4 was similar. These results indicated that C/EBP specifically dimerized in vitro with itself and ATF4, and only the bZIP domain of ATF4 was required.
|
pulled down full-length C/EBP, ATF4, or ATF4249–351 (Fig. 2B
, Lanes 2 and 4). In contrast, the GST-C/EBP147–249 fusion protein pulled down full-length ATF4, but not C/EBP30 or ATF4249–351 (Fig. 2B
, Lane 3). An extended exposure of the gel revealed very weak binding of the GST-C/EBP147–249 fusion with C/EBP30 (data not shown). The GST-ATF425–351 fusion efficiently pulled down C/EBP30 and the ATF4 and ATF4249–351 proteins (Fig. 2B
, Lane 5). These results demonstrated that the bZIP domain of C/EBP was required for interaction with itself and ATF4. In addition, C/EBP only dimerized efficiently in vitro when one of the partners contained sequences N-terminal to the bZIP domain. These results further supported the observation that amino acid residues outside of the ZIP domain enhanced homo- and heterodimerization of C/EBP.
To determine the extent of C/EBP interaction with CREB family members, pull-down assays with GST fusions of CREB1 and CRE-BP1/ATF2 were performed. The GST-CREB1 fusion protein interacted with C/EBP, but not with either form of ATF4 (Fig. 2B
, Lane 6). The GST-ATF2 fusion protein pulled down C/EBP and both forms of ATF4 (Fig. 2B
, Lane 7). These results suggested that C/EBP may heterodimerize with other CREB family members in addition to ATF4. These other CREB/ATF family members were not isolated during the yeast two-hybrid screen; therefore, we concentrated on the ATF4:C/EBP interaction for the remainder of this study.
To determine if the C/EBP:ATF4 interaction occurred in cells, expression vectors encoding the transcriptionally active isoforms of C/EBP, -32, and -30 were transfected alone or with an ATF4 expression vector into COS-1 cells (Fig. 2C)
. Lysates were immunoprecipitated with anti-ATF4 antiserum, electrophoresed through a polyacrylamide gel, and analyzed by immunoblot using a polyclonal anti-C/EBP antibody. Both isoforms were coimmunoprecipitated specifically and reproducibly by the ATF4 antibody (Fig. 2C
, Lanes 6 and 8). This was not observed in lysates prepared from cells transfected with empty expression vector (Fig. 2C
, Lane 4) or individual expression vectors (Fig. 2C
, Lanes 5, 7, and 9). Taken together, our results showed that the interaction between C/EBP and ATF4 occurred in vitro, in yeast, and in mammalian cells.
ATF4:C/EBP heterodimers bind to asymmetric C/EBP sites
Heterodimerization between the CREB/ATF and other bZIP family members redirects their binding to DNA regulatory elements [45
, 47
, 52
]. To determine the effect of ATF4 on C/EBP DNA-binding properties, EMSAs, using previously defined C/EBP-binding site oligonucleotides and purified GST-ATF425–351 and MBP-C/EBP30 fusion proteins were performed (results summarized in Table 2
). A symmetric (palindromic core motif) consensus C/EBP site was bound by C/EBP but not by ATF4 (Fig. 3A
, top panel, Lanes 2 and 4). This binding was specific, as it was competed by a 100-fold excess of unlabeled "self" but not a nonspecific oligonucleotide (Fig. 3A
, Lanes 9 and 10). Increasing the concentration of C/EBP protein in the reaction increased homodimer complex formation (data not shown). In contrast, when both proteins were incubated (1:1 molar ratio) with the oligonucleotide, complex formation increased (Fig. 3A
, Lane 3). The complex was supershifted by antibodies against C/EBP or ATF4 (Fig. 3A
, Lanes 5 and 6). Unlike the homodimer, the heterodimer bound nonspecifically. Excess cold self and nonspecific oligonucleotides were unable to compete for binding to the DNA (Fig. 3A
, Lanes 7 and 8). These results indicated that C/EBP homodimers specifically bound the consensus C/EBP site, but ATF4:C/EBP heterodimers did not.
|
|
homodimer bound to both sites (Fig. 3A
, Lane 2); however, the G-CSFR site was bound less efficiently than the NE site. Binding to both sites was competed specifically by excess oligonucleotide (Fig. 3A
, Lanes 9 and 10). When both proteins were mixed, the amount of G-CSFR oligonucleotide shifted increased dramatically (Fig. 3A
, middle panel, Lane 3). The amount of NE probe shifted did not show a significant increase (Fig. 3A
, bottom panel, Lane 3). Antibody against C/EBP or ATF4 supershifted complexes for the NE and G-CSFR sites, indicating that a heterodimer of both proteins was bound (Fig. 3A
, middle and bottom panels, Lanes 5 and 6). The proportion of the G-CSFR complex supershifted by the anti-ATF4 antibody was significantly higher than for the NE complex (Fig. 3A
, Lane 6, middle and bottom panels, respectively). The heterodimer binding was specific, as 100-fold excess cold self but not the nonspecific oligonucleotide abrogated the complex formation (Fig. 3A
, Lanes 7 and 8). The results from the above experiments, which were repeated two to three times for each oligonucleotide, indicated that ATF4:C/EBP heterodimers bound specifically to asymmetric but not symmetric C/EBP sites.
ATF4:C/EBP heterodimers bind to asymmetric CRE sites
Previous studies demonstrated that another cross-family interaction involving C/EBPß and C/ATF (murine ATF4) directed binding of the heterodimers to CRE sites rather than C/EBP sites [45
]. The symmetric consensus CRE site present in the SOM promoter was bound by homodimers of ATF4 but not C/EBP (Fig. 3B
, top panel, Lane 4 vs. 2). The binding was competed specifically by excess cold oligonucleotides (Fig. 3B
, Lanes 7 and 8) and supershifted by anti-ATF4 but not anti-C/EBP antibody (Fig. 3B
, Lanes 5 and 6). These results indicated that the heterodimer does not bind to the consensus CRE site. Similar results were observed for the symmetric consensus site present in the COL-8 gene (Table 2)
.
The asymmetric CRE site in the ENK gene was bound weakly by C/EBP homodimers (Fig. 3B
, middle panel, Lane 2). Binding was specific, as determined by competition with excess cold oligonucleotides (Fig. 3B
, Lanes 9 and 10). In contrast, ATF4 homodimers were unable to bind the ENK site (Fig. 3B
, Lane 4). When both proteins were mixed, binding increased significantly (Fig. 3B
, Lane 3). Incubation with excess cold oligonucleotide demonstrated binding was specific (Fig. 3B
, Lanes 7 and 8). Both antibodies supershifted the complex, indicating binding by a heterodimer (Fig. 3B
, Lanes 5 and 6). Similar results were observed for the PEPCK asymmetric CRE site (Table 2)
and a synthetic chimeric site composed of half sites from the consensus CRE and C/EBP sites (Fig. 3B
, bottom panel).
These results demonstrated that heterodimerization of ATF4 with C/EBP redirected both proteins to bind efficiently to asymmetric rather than symmetric C/EBP or CRE sites, which are preferred by the homodimers. It is interesting that at higher protein concentrations, C/EBP homodimers bound to most of the sites tested (data not shown); however, ATF4 homodimers only bound efficiently to the symmetric CRE sites, SOM and COL-8. A consensus-binding site, 5'-TGACGCAA-3', was derived by compiling the core motifs from each oligonucleotide bound by the heterodimer. It resembled the synthetic, chimeric-binding site (Table 2)
.
ATF4 affects transcription from myeloid-specific gene promoters that contain C/EBP-binding sites
The EMSA studies suggested that ATF4 might affect transcriptional activation from promoters that contain C/EBP sites. To examine this, combinations of c-Myb, C/EBP, and ATF4 expression vectors were cotransfected into the Jurkat T cell line with the pMim-Luc reporter containing the HindIII (–242) to XhoI (in the first intron) fragment of the mim-1 gene [33
, 34
]. Low luciferase activity was detected in cells cotransfected with pMim-Luc and an empty or c-Myb expression vector (Fig. 4A
and 4B
). In the absence of c-Myb, C/EBP32, and C/EBP activated the reporter similarly with five- to sixfold less activation by C/EBP30 (Fig. 4A)
. Each C/EBP protein synergistically activated transcription with c-Myb (Fig. 4B)
. In the presence of c-Myb, C/EBP was consistently a much stronger activator of transcription than either C/EBP isoform (Fig. 4B)
. C/EBP plus c-Myb activated the promoter 60-fold above c-Myb alone, whereas cotransfection of C/EBP32 or C/EBP30 with c-Myb resulted in a 20- and tenfold increase, respectively (Fig. 4B)
. Cotransfection of ATF4 with c-Myb resulted in a threefold increase in activity over c-Myb alone (Fig. 4B)
. ATF4 did not activate transcription significantly from pMim-Luc in the absence of c-Myb, and it had no obvious effect on C/EBP-mediated transcription (Fig. 4A)
. In contrast, ATF4 decreased C/EBP-mediated transcription by 80% (Fig. 4A)
. It is surprising that when ATF4 was cotransfected with c-Myb plus C/EBP32 or -30, promoter activity jumped dramatically to 100- and 140-fold, respectively, over c-Myb alone (Fig. 4B)
. In contrast, activation by C/EBP dropped 25–30% to 45-fold when cotransfected with c-Myb and ATF4 (Fig. 4B)
. The synergistic activation of the mim-1 promoter by C/EBP, c-Myb, and ATF4 responded in a dose-dependent manner with increasing ATF4 expression vector but leveled off with ratios higher than 1:1 (Fig. 4C)
. Increasing the amount of ATF4 expression vector decreased the activation by C/EBP at ratios of 1:1 and 3:1 ATF4:C/EBP (Fig. 4C)
. These results indicated the activating and inhibitory effects of ATF4 were dose-dependent.
|
32 consistently activated three- to fourfold better than C/EBP30 (Fig. 4A)
. The additional 32 N-terminal amino acid residues in C/EBP32 make it a more potent transcriptional activator than C/EBP30 [33
]. When ATF4 was cotransfected with either C/EBP isoform, in the presence of c-Myb, the differences in activation were not evident (Fig. 4B)
, suggesting that the TAD of ATF4 is critical for this synergistic increase. Transfection of a truncated ATF4 lacking the TAD (ATF4249–351) resulted in repression of C/EBP30- and C/EBP-mediated transcription (Fig. 4D)
. The repression of C/EBP-mediated transcription was similar to that observed with the full-length ATF4 (Fig. 4D)
. In contrast, removal of the TAD of C/EBP (C/EBP85–249) abolished C/EBP cooperation with c-Myb and muted (fivefold decrease) but did not abolish the transcriptional synergy among C/EBP, c-Myb, and ATF4 (Fig. 4D)
. These results indicated that the TAD of ATF4 but not C/EBP is essential for synergistic activation, and the bZIP domain of ATF4 is sufficient for repression of C/EBP-mediated transcription.
Previous studies have shown that cooperative activation of the mim-1 promoter by C/EBP proteins requires c- or v-Myb [7
, 31
]. Similarly, the synergistic activation of the mim-1 promoter by ATF4 requires coexpression of c-Myb (Fig. 4)
. C/EBP-mediated transcriptional activation is inhibited more severely by ATF4 in the absence of c-Myb (25% vs. 80% reduction; Fig. 4
, panel A vs. B).
The C/EBP:ATF4 heterodimer binds a previously uncharacterized C/EBP site in the mim-1 promoter
Two C/EBP sites are present in the mim-1 promoter, one at –60 (symmetric) and the other at –160 (asymmetric) from the transcriptional start site [7
]. To determine if either were required for synergistic activation by the heterodimer, we performed experiments with reporter constructs containing a mutation in one or the other site [7
]. Mutation of the C/EBP site at –60 (M60-Luc) did not alter the overall activity of the promoter or its response to the C/EBP30:ATF4 heterodimer with c-Myb (Fig. 5A
). Mutation of the C/EBP site at –160 (M160-Luc) resulted in a reduction of the overall activity of the reporter compared with the wild-type or M60-Luc, but the synergistic activation with the C/EBP:ATF4 heterodimer and c-myb was intact (Fig. 5A)
. Both mutated promoters were activated synergistically by the heterodimer such as the wild-type promoter (Fig. 5A)
. This suggested that neither site was essential for the synergistic activation by the C/EBP:ATF4 heterodimer. This prediction was supported by EMSA experiments using double-stranded oligonucleotides (Mim-60 and Mim-160), representing each site (Fig. 5B)
. Only C/EBP homodimers bound specifically to these sites, and the heterodimer was unable to bind (Fig. 5B
, top and middle panels, Lanes 2 and 5–10). In fact, ATF4 reduced the binding of the C/EBP homodimers to Mim-60 and Mim-160 (Fig. 5B
, Lane 3). We tested other computer-predicted, C/EBP-binding sites and found that one located at nucleotide –174 (Mim-174) was bound specifically by C/EBP homodimers and C/EBP:ATF4 heterodimers (Fig. 5B
, bottom panel, Lanes 2, 3, and 5–10). The ATF4 protein did not bind to any of the sites as a homodimer (Fig. 5B
, Lane 4). Our results suggest this third binding site in the mim-l promoter at nt –174, which overlaps the Myb-binding box B, may be involved in transcriptional activation by the heterodimer [46
].
|
:ATF4 heterodimer synergistically activates promoters for other myeloid-specific genes, the NE-Luc promoter was activated synergistically by the addition of ATF4 (Fig. 6A)
. The effect was most pronounced for C/EBP30 (five- vs. twofold for C/EBP32). Again, activation by C/EBP was inhibited 25% (Fig. 6A)
. When these same experiments were performed with a NE-Luc reporter construct containing a mutant C/EBP site or lacking the NE promoter, transcriptional activation was not observed (data not shown). This indicated that a functional C/EBP site was required for the C/EBP:ATF4 heterodimer to bind to the promoter and agrees with the EMSA results, which showed the heterodimer bound this site.
|
:ATF4 heterodimer. It is unexpected that the C/EBP-mediated activation was similar in the presence or absence of ATF4; however, C/EBP activation was inhibited by 50% (Fig. 6B)
. Activity was not observed with a reporter lacking the G-CSFR promoter (data not shown). These results resembled those with the pMim-Luc reporter when c-Myb was excluded (Fig. 6A)
. It appears that homodimers of C/EBP and heterodimers with ATF4 activate the G-CSFR promoter equally well, whereas C/EBP-mediated transcription is inhibited upon heterodimerization with ATF4.
ATF4 interacts with C/EBP to form a transcriptionally less-active dimer
For C/EBP, the mechanism of inhibition is unclear but may involve decreased binding by the C/EBP:ATF4 heterodimer to the C/EBP sites present in the promoters tested, or the heterodimer is transcriptionally less active than the C/EBP homodimer. To test these possibilities, we determined if C/EBP would heterodimerize with ATF4 and if so, what effect this had on DNA binding. Pull-down assays using GST, GST-ATF4, GST-ATF2, and GST-CREB1 fusion proteins and in vitro-synthesized C/EBP labeled with 35S-methionine were performed (Fig. 7A
). ATF2 was shown to heterodimerize with C/EBP and was included as a positive control [47
]. As expected, ATF2 pulled down C/EBP, and GST did not (Fig. 7A
, Lanes 4 and 1, respectively). In addition, ATF4 and CREB1 pulled down C/EBP (Fig. 7A
, Lanes 2 and 3, respectively). These results demonstrated that C/EBP dimerizes with ATF/CREB family members including ATF4.
|
and -, we performed EMSAs using the NE and G-CSFR C/EBP sites with an increasing dose of ATF4 (Fig. 7B)
. The GST-C/EBP and MBP-C/EBP fusion proteins bound to each site as a homodimer (Fig. 7B
, Lanes 1 and 6). For the NE site, increasing the molar concentration of ATF4 decreased C/EBP homodimer binding by 50% at an 8:1 ratio (Fig. 7B
, Lanes 2-4). In contrast, C/EBP binding was relatively unaffected (Fig. 7B
, Lanes 7–9). As expected, addition of anti-ATF4 antibody supershifted C/EBP:ATF4 and C/EBP:ATF4 heterodimer complexes (Fig. 7B
, Lanes 10 and 5, respectively). For the G-CSFR site, increasing the molar concentration of ATF4 resulted in decreased C/EBP homodimer binding with a concomitant increase in heterodimer binding (Fig. 7B
, Lanes 1–5). It is interesting that C/EBP homodimer binding was relatively unaffected, but heterodimer binding increased with increasing ATF4 (Fig. 7B
, Lanes 6–10). The presence of the heterodimer was demonstrated by supershifting the faster migrating complex with anti-ATF4 antibody (Fig. 7B
, Lanes 5 and 10). The C/EBP sites from the mim-1 (Mim-60 and Mim-160) promoter demonstrated a decrease in C/EBP binding at high concentrations of ATF4, similar to that for the NE site; however, heterodimer binding was not observed (data not shown). In addition, we noted binding of C/EBP:ATF4 heterodimers to the chimeric site, which was similar to the binding observed for the G-CSFR site (data not shown).
Taken together, the data suggest that although C/EBP homodimer formation decreases, C/EBP:ATF4 dimers form and bind to the same sites as C/EBP:ATF4 heterodimers. Decreased C/EBP homodimer formation and binding may account for some of the inhibition observed in the transfections, especially for NE and mim-1. However, the similar binding patterns for the NE and G-CSFR C/EBP sites and inhibition of C/EBP-mediated transcription from these promoters suggest that interaction of C/EBP with ATF4 creates a transcriptionally less active dimmer, which replaces the more active C/EBP homodimer.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
and -ß interacted with ATF2 [47
]. In addition, CREB-1 interacts with C/EBPß to activate transcription from the human pro-IL-1ß gene, and ATF3 transcriptional activity is modulated by C/EBP
/GADD153/CHOP 10 [53
, 54
]. Finally, mouse C/ATF dimerizes with C/EBPß, -, and - [45
, 55
]. In addition to ATF4, our in vitro, pull-down assays suggest that C/EBP potentially interacts with CREB1 and ATF2 (Fig. 2B)
. The overwhelming isolation of ATF4 clones in the screen with the yeast two-hybrid system indicated that it is a potentially important dimerization partner for C/EBP.
Cross-family dimerization involving ATF4 alters the DNA-binding specificities of both partners. The ATF4 protein dimerizes with Fra-1, Fos, and Jun in vitro and directs binding specificities to symmetric CRE sites not bound efficiently by either homodimer [52
]. When C/ATF dimerizes in vitro with C/EBPß or -, it directs binding to symmetric (SOM) and asymmetric (ENK and PEPCK) CRE sites [45
]. The asymmetric sites are not efficiently bound by either homodimer. In contrast, the ATF2:C/EBP heterodimer did not bind to the consensus CRE site [47
]. Although most of these heterodimers bound to symmetric or asymmetric CRE sites, they did not bind to the C/EBP sites tested [45
, 47
]. We found that ATF4:C/EBP heterodimers bind preferentially to asymmetric but not symmetric C/EBP or CRE sites. Binding of the ATF4:C/EBP heterodimer to symmetric CRE sites was observed at higher concentrations of protein (data not shown); however, under conditions where specific binding to asymmetric sites was observed, specific binding to the symmetric CRE sites did not occur (Fig. 3B)
. As in the other studies, heterodimerization allows ATF4 to bind to sites that it could not bind as a homodimer, thereby expanding the variety of CRE and C/EBP sites it may potentially regulate. As shown for ATF2:C/EBP and aplysia ATF4:C/EBP heterodimers, the ATF4:C/EBP dimers also bind efficiently to an artificially constructed hybrid CRE-C/EBP (chimeric) site (Table 2)
[47
, 56
]. A compilation of the natural sites tested in this study indicates that the heterodimer-binding consensus core motif is identical to this chimeric site (Table 2)
.
ATF4 represses CREB1 and ATF2-mediated transcription [38
, 56
, 57
]; however, it cooperates with Tax to activate the human T cell lymphotropic virus 1-long-terminal repeat. In addition, ATF4 interacts with the CREB-binding protein (CBP) and activates transcription from CRE-containing reporters [58
]. Based on the EMSA results, we predicted that ATF4 would affect the regulation of genes, which are activated transcriptionally by C/EBP proteins. The C/EBP protein was included initially as a control, as it is generally a stronger transcriptional activator of early myeloid-specific genes than C/EBP (Fig. 4)
[15
]. Cotransfection of ATF4 with C/EBP results in a consistent and significant decrease in transcriptional activity with all promoters tested as compared with C/EBP alone. It is surprising that cotransfection of ATF4 and C/EBP resulted in an extremely potent increase of gene transcription as compared with C/EBP alone. This was evident, particularly with promoters that are cooperatively activated by C/EBP and c-Myb proteins (e.g., Mim-1 and NE). The levels were equivalent to or higher than those by C/EBP homodimers.
For C/EBP, transcription from the G-CSFR promoter was relatively unaffected by the absence or presence of ATF4; however, C/EBP-mediated transcription decreased 50% when ATF4 was present. One possible explanation is that C/EBP:ATF4 heterodimers did not form in the cells and bind to the G-CSFR promoter. Although we have not ruled this out, the gel-shift data clearly demonstrate that the heterodimer can bind to the C/EBP site in the G-CSFR promoter. In addition, the cotransfection of C/EBP and ATF4 leads to synergistic activation of the NE and Mim-1 promoters, indicating the presence of a heterodimer in the cells. We propose that an additional transcription factor may be required to interact with the C/EBP:ATF4 heterodimer to confer synergistc activation of the G-CSFR promoter, much as c-myb does with the Mim-1 and NE promoters. Such factors may be absent in the Jurkat T cell line. We tested one such factor, PU.1, but found that it did not cooperatively activate the G-CSFR gene with C/EBP or - (data not shown). Alternatively, the G-CSFR gene may not be a target of the C/EBP:ATF heterodimer and therefore, does not possess the sites to bind other factors, which would cooperate with the heterodimer. The physiological relevance of this study remains to be addressed. We have identified ATF4 as a heterodimerization partner for C/EBP and - and used promoter constructs for previously identified C/EBP target genes to demonstrate that heterodimerization results in mechanistically interesting results. However, it remains to be determined what the actual in vivo target genes for these heterodimers are. To this end, studies involving cloning of chromatin IP (ChIP) products and microarray analysis of ChIP roducts (ChIP-on-chip) are underway in myeloid cells, in which the balance between these proteins is shifted to identify physiologically important target genes.
The dramatic increase in transcriptional activation by the C/EBP:ATF4 heterodimer appears to involve two mechanisms: The heterodimer binds with higher affinity than either homodimer (Fig. 3A)
, and the ATF4 TAD dramatically increases the transcriptional activity of the heterodimer. This was especially evident when the TAD of C/EBP was deleted, and the heterodimer still activated transcription as well as the full-length C/EBP homodimer (Fig. 4D)
. The coactivator proteins CBP and p300 interact with ATF4 [58
], c-Myb [59
, 60
], and C/EBP family members [27
]. We have not detected an interaction between C/EBP and CBP/p300 (Walter Verbeek and H. P. Koeffler, unpublished observation) and therefore, hypothesize that a heterodimer of C/EBP and ATF4 would interact with and recruit CBP/p300 to the promoter more efficiently than a homodimer of C/EBP. The dramatic increase in transcription from promoters containing C/EBP and Myb-binding sites compared with those lacking Myb-binding sites may result from more efficient recruitment of coactivators to the promoter by ATF4 and c-Myb, which are capable of interacting with CBP/p300. Consistent with this, CBP potentiates the synergistic transcription mediated by c- or v-Myb with NF-M from the mim-1 promoter [59
]. The synergy between the C/EBP:ATF4 heterodimer and c-Myb may involve formation of a stable complex between ATF4 and c-Myb; however, such an interaction has not been reported, and a direct interaction between C/EBP and c-Myb has not been observed, although these two factors synergistically activate gene transcription [7
]. The inhibition of C/EBP-mediated transcription appears to involve decreased C/EBP homodimer binding and the formation of C/EBP:ATF4 heterodimers, which bind DNA but do not activate transcription as well as C/EBP homodimers. We hypothesize that the TAD of ATF4 enhances the usually weak C/EBP transcriptional activity but inhibits the usually potent C/EBP activity.
The differential regulation of myeloid gene expression by ATF4 via interactions with C/EBP and C/EBP has particular importance in myelopoiesis, as C/EBP and C/EBP are expressed in an overlapping manner in myeloid cells. All C/EBP family members bind to the same C/EBP consensus site [5
]. Regulation of different sets of genes by C/EBP proteins is unclear. It is interesting that Khanna-Gupta et al. [61
] demonstrated that C/EBP may transcriptionally repress primary and secondary granule genes, and C/EBP activates those same genes during neutrophil differentiation, concomitant with an increase in its own expression levels and binding to target genes, which are repressed by C/EBP. They proposed that a yet-unidentified partner for C/EBP may be involved in the repression. We postulate that ATF4 could serve to repress and activate the same gene(s) depending on the heterodimerization partner (C/EBP:ATF4, repression; C/EBP:ATF4, activation).
Recent studies place ATF4 at the intersection of multiple intracellular stress pathways including unfolded protein response, amino acid starvation, and oxidative stress [62
, 63
]. These stress signals result in phosphorylation of eukaryotic initiation factor-2, which produces a general inhibition of translation but increased translation of specific mRNAs including ATF4 [64
], which up-regulates genes involved in antioxidant and amino acid metabolism [62
]. The ATF4 protein was found up-regulated by hypoxic conditions [65
66
67
], which prolong the survival of neutrophils [68
]. It has not been shown that hypoxic conditions induce ATF4 or C/EBP in neutrophils; however, we propose that ATF4:C/EBP heterodimers could play a role in regulating genes, which are induced during exposure of neutrophils to hypoxic conditions at sites of infection of inflammation [69
]. It is interesting that studies demonstrate that anoxic conditions induce C/EBPß and ATF4 in fibroblasts [65
], supporting a possible role for these two families in regulating important innate immune responses.
|
ACKNOWLEDGEMENTS |
|---|
Received August 15, 2006; revised January 17, 2007; accepted February 6, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
on the serum albumin promoter Genes Dev. 8,350-362 J. Biol. Chem. 273,14796-14804-1 acid glycoprotein gene Mol. Cell. Biol. 11,4959-4965-deficient mice Proc. Natl. Acad. Sci. USA 94,569-574) Blood 90,2987-2994, is upregulated during granulocytic, but not monocytic, differentiation Blood 90,2591-2600 is preferentially up-regulated during granulocytic differentiation and its functional versatility is determined by alternative use of promoters and differential splicing Proc. Natl. Acad. Sci. USA 94,6462-6467-deficient mice Proc. Natl. Acad. Sci. USA 94,13187-13192 J. Exp. Med. 189,1847-1852 Blood 97,2561-2567 (C/EBP), in acute myeloid leukemia Nat. Genet. 27,263-270[CrossRef][Medline] in myelodysplastic syndromes and acute myeloid leukemias Blood 99,1332-1340
