Double Plant Homeodomain (PHD) Finger Proteins DPF3a and -3b Are Required as Transcriptional Co-activators in SWI/SNF Complex-dependent Activation of NF-κB RelA/p50 Heterodimer

Background: The NF-κB dimer, RelA/p50, often requires the SWI/SNF complex for its transactivation function, but its molecular mechanisms remain elusive.

Results: The NF-κB canonical pathway induced by TNF-α is DPF3a/b- and SWI/SNF-dependent for some promoters.

Conclusion: DPF3a and DPF3b are effective linkers for the SWI/SNF complex and RelA/p50.

Significance: The NF-κB·DPF3a/b·SWI/SNF complex would be an effective platform for promoter-specific transactivation.

Abstract

We have previously shown that DPF2 (requiem/REQ) functions as a linker protein between the SWI/SNF complex and RelB/p52 NF-κB heterodimer and plays important roles in NF-κB transactivation via its noncanonical pathway. Using sensitive 293FT reporter cell clones that had integrated a SWI/SNF-dependent NF-κB reporter gene, we find in this study that the overexpression of DPF1, DPF2, DPF3a, DPF3b, and PHF10 significantly potentiates the transactivating activity of typical NF-κB dimers. Knockdown analysis using 293FT reporter cells that endogenously express these five proteins at low levels clearly showed that DPF3a and DPF3b, which are produced from the DPF3 gene by alternative splicing, are the most critical for the RelA/p50 NF-κB heterodimer transactivation induced by TNF-α stimulation. Our data further show that this transactivation requires the SWI/SNF complex. DPF3a and DPF3b are additionally shown to interact directly with RelA, p50, and several subunits of the SWI/SNF complex in vitro and to be co-immunoprecipitated with RelA/p50 and the SWI/SNF complex from the nuclear fractions of cells treated with TNF-α. In ChIP experiments, we further found that endogenous DPF3a/b and the SWI/SNF complex are continuously present on HIV-1 LTR, whereas the kinetics of RelA/p50 recruitment after TNF-α treatment correlate well with the viral transcriptional activation levels. Additionally, re-ChIP experiments showed DPF3a/b and the SWI/SNF complex associate with RelA on the endogenous IL-6 promoter after TNF-α treatment. In conclusion, our present data indicate that by linking RelA/p50 to the SWI/SNF complex, DPF3a/b induces the transactivation of NF-κB target gene promoters in relatively inactive chromatin contexts.

Introduction

NF-κB³ is a key transcription factor that regulates many biological processes, such as immune, inflammatory, and virus responses, development, cellular growth, and apoptosis. The regulation by NF-κB is achieved through the transactivation of a large number of its target genes in a cell type-specific and/or stimulus-specific manner (1–4). NF-κB is itself composed of homo- or heterodimeric complexes of members of the NF-κB family of proteins, which include RelA (p65), RelB, c-Rel, p50, and p52 in humans. NF-κB dimers remain inactive in the cytoplasm until specific stimulation activates their signaling pathways (5). One of the main NF-κB pathways, the canonical pathway, is triggered by stimulation with factors such as tumor necrosis factor-α (TNF-α) and lipopolysaccharide (LPS). The induction of this pathway results in the activation of the RelA/p50 heterodimer by transporting it to the nucleus after phosphorylation and also the proteasomal degradation of inhibitor of NF-κB (IκB), which retains RelA/p50 in the cytoplasm under unstimulated conditions (6). However, RelB/p52, which is present in the cytoplasm as inactive RelB/p100 (NF-κB1) until stimulated, is activated through the noncanonical pathway (7). This pathway is triggered by stimuli such as lymphotoxin exposure, and it induces the cleavage of cytosolic p100 to produce p52, which subsequently translocates into the nucleus as a RelB/p52 dimer.

The mechanism by which each NF-κB dimer is recruited to a certain set of promoters and specifically transactivates their transcription in a cell type-dependent manner is an important question. Such processes would likely involve chromatin context-dependent regulation. Indeed, NF-κB transactivation is known to often require the SWI/SNF complex, a representative chromatin remodeling factor involved in epigenetic regulation in humans, but no clear and direct interaction between NF-κB components and the SWI/SNF complex has been reported. The SWI/SNF complex has two alternative ATPases, Brahma (Brm) or Brahma-related gene 1 (BRG1) as the catalytic subunits, and other subunits such as BAF155, Ini1/SNF5, BAF170, BAF60a, and β-actin (8–10). The SWI/SNF complex is recruited to target genes via association with transcription regulators such as c-Myc (11), C/EBPβ (12), AP-1 (13), neuron-restrictive silencer factor (14), and Cdx2 (15). In addition to cellular targets, we have shown that integrated LTRs of murine leukemia virus (16, 17) and HIV-1 (human immunodeficiency virus-1) (18) require the Brm-type SWI/SNF complex to maintain their gene expression.

We have recently shown that DPF2 (REQ/BAF45d) functions as an efficient adaptor protein between the SWI/SNF complex and RelB/p52 and plays important roles in NF-κB transcriptional activation at the most downstream part of the NF-κB noncanonical pathway (19). DPF2 belongs to the d4 family of proteins, the members of which are characterized by an N-terminal 2/3 domain containing a nuclear localization signal, a central C2H2-type Krüppel-like zinc finger motif, and a C-terminal d4 domain consisting of a tandem repeat of the PHD zinc finger (20–22). It was previously shown that another member of the d4 family, DPF3 (Cerd4/BAF45c), associates with the SWI/SNF complex and further that the DPF3 gene has two splicing variants, the products of which differ at their C-terminal regions. The DPF3b protein shows all of the characteristics of a d4 family member, including binding activity to either methylated or acetylated lysine residues at histones H3 and H4 through the PHD fingers. However, DPF3a lacks these binding activities as it harbors a truncated d4 domain within its first PHD (23). Interestingly, two d4 family members, DPF1 (Neud4/BAF45b) and DPF3, and an additional protein, PHF10 (BAF45a), which also possesses double PHD fingers within its C terminus, have been reported to associate with the SWI/SNF complex in neural cells in a differentiation-specific manner (24). In mouse neural progenitors, the complex contains PHF10, but this component is substituted by DPF1 or DPF3 when the cells become differentiated into post-mitotic neurons.

Considering our previous finding that DPF2 is required for RelB/p52 transactivation via the SWI/SNF complex, we speculate that the five proteins DPF1, DPF2, DPF3a, DPF3b, and PHF10 are candidate co-activators of the typical NF-κB heterodimer, RelA/p50, as well as two other NF-κB dimers, RelB/p52 and c-Rel/p50. We show in our current analysis that each of these proteins can enhance the different NF-κB heterodimers to transactivate their targets efficiently when both they and the NF-κB components are exogenously expressed. We further show from our analysis that among these five proteins DPF3a and DPF3b are the most effective cofactors for RelA/p50 activation in 293FT cells treated with TNF-α. Our current data further indicate that these two proteins directly bind to RelA, p50, and at least four subunits of the SWI/SNF complex in vitro. We additionally show that endogenous DPF3a/b and the SWI/SNF complex are continuously co-localized at the HIV-1 LTR throughout the period of TNF-α stimulation and that RelA/p50 is promptly recruited to the typical NF-κB-binding sites within the HIV-1 LTR or endogenous IL-6 promoter upon stimulation. Finally, we show that the viral transcripts are synthesized with similar kinetics to those of RelA/p50 recruitment.

EXPERIMENTAL PROCEDURES

Cell Culture and Retro- and Lentiviral Vectors

293FT cells (Invitrogen) were maintained in Dulbecco's modified Eagle's medium (Wako Chemicals, Tokyo) containing 10% fetal calf serum at 37 °C and 5% CO₂. For TNF-α stimulation, the cells were treated with 10 ng/ml TNF-α (R&D Systems, Minneapolis, MN). Vesicular stomatitis virus G protein (VSV-G)-pseudotyped retro- or lentiviral vectors were prepared as described previously (17, 19). For transduction, cells were incubated with the virus vector stocks in the presence of 8 μg/ml Sequa-brene^TM (Sigma).

Antibodies

Rabbit polyclonal antibodies against human DPF3 were raised against a synthetic peptide corresponding to amino acid residues 230–251 of the protein (NP_036206) by Medical and Biological Laboratories (MBL, Nagoya, Japan). Other antibodies used in this research are as follows: normal rabbit IgG (PM035) (MBL), α-RelA (C-20), α-BRG-1 (H-88), α-BAF155 (H-76) (Santa Cruz Biotechnology, Santa Cruz, CA), α-p105/p50 (ab7971), α-Brm (ab15597) (Abcam, Cambridge, MA), α-FLAG (M2) (Sigma), α-BAF60a (611728), and α-Ini1 (612110) (BD Transduction Laboratories, San Jose, CA). Most of these specific antibodies are rabbit polyclonal, except α-FLAG, α-BAF60a, and α-Ini1 which are mouse monoclonal.

GST Fusion Protein Pulldown Assays

GST fusion proteins were expressed in Escherichia coli Rosetta 2 (Novagen, Madison, WI) via incubation of the cells with 0.1 mm isopropyl 1-thio-β-d-galactopyranoside overnight at 15 °C and then prepared using His tag Binding/Wash Buffer (20 mm Tris-HCl (pH 8.0), 600 mm NaCl, 1 mm MgCl₂, 10% glycerol, 0.1% Nonidet P-40, 20 mm imidazole, and phosphatase inhibitors). Purification of GST-His-tagged proteins was performed sequentially using Profinity IMAC nickel-charged resin (Bio-Rad) and glutathione-Sepharose 4B (GE Healthcare). The target proteins were eluted from these columns with His tag Elution Buffer (20 mm Tris-HCl (pH 8.0), 600 mm NaCl, 1 mm MgCl₂, 10% glycerol, 0.1% Nonidet P-40, 250 mm imidazole, and phosphatase inhibitors) and GST Elution Buffer (100 mm Tris-HCl (pH 8.0), 12 mm NaCl, 20 mm glutathione, and phosphatase inhibitors), respectively. In vitro transcription and translation were then performed as described previously (19). In vitro synthesized proteins were incubated with respective GST fusion proteins for 2 h at 4 °C on a rotating platform. The complexes were washed for five times with TNE buffer (10 mm Tris-HCl (pH 7.8), 150 mm NaCl, 1 mm EDTA, 1% Nonidet P-40, and protease inhibitors) and analyzed by SDS-PAGE followed by autoradiography using FLA-5100 (Fujifilm). Cellular lysates, prepared in TN buffer (10 mm Tris-HCl (pH 7.8), 150 mm NaCl, 1% Nonidet P-40, protease inhibitors, and phosphatase inhibitors) were treated with 100 units/ml benzonase endonuclease (Novagen) and precleared by incubation with GST protein immobilized on beads for 1 h. The lysate was then incubated with respective GST fusion proteins for 4 h at 4 °C on a rotating platform. The complexes were subsequently washed three times with Buffer D (20 mm HEPES-KOH (pH 7.9), 20% glycerol, 0.1 m KCl, 0.2 mm EDTA, 0.1% Nonidet P-40, and protease inhibitors) and analyzed by immunoblotting.

Chromatin Immunoprecipitation (ChIP) and Sequential ChIP Assays

Cells were cross-linked with 1% formaldehyde at 37 °C for 8 min. Cross-linking reactions were stopped by the addition of a 0.1 volume of 1.5 m glycine and incubation at room temperature for 5 min. The cells were then washed with PBS, collected, and incubated in lysis buffer 1 (50 mm HEPES (pH 7.5), 140 mm NaCl, 1 mm EDTA, 10% glycerol, 0.5% Nonidet P-40, 0.25% Triton X-100, and protease inhibitors). Cell lysates were sequentially replaced with lysis buffer 2 (10 mm Tris-HCl (pH 8.0), 200 mm NaCl, 1 mm EDTA, 0.5 mm EGTA, and protease inhibitors), and lysis buffer 3 (10 mm Tris-HCl (pH 8.0), 300 mm NaCl, 1 mm EDTA, 0.5 mm EGTA, 0.1% sodium deoxycholate, 0.5% N-lauroylsarcosine, and protease inhibitors). The lysates were then sonicated using Elestein Ryuseikai (Elekon Science. Corp., Chiba, Japan) on ice so that the DNA would be sheared into small fragments with an average length of less than 0.5 kb. After the addition of a 0.1 volume of 10% Triton X-100, the lysates were centrifuged to remove cellular debris and incubated overnight on a rotating platform at 4 °C with the respective antibodies (10 μg of each), which were previously bound to Dynabeads protein G (Invitrogen). The beads were washed once in low salt buffer (20 mm Tris-HCl (pH 8.0), 150 mm NaCl, 2 mm EDTA, 0.1% SDS, 1% Triton X-100, and protease inhibitors), twice in high salt buffer (20 mm Tris-HCl (pH 8.0), 400 mm NaCl, 2 mm EDTA, 0.1% SDS, 1% Triton X-100, and protease inhibitors), five times in RIPA buffer (50 mm HEPES (pH 7.6), 500 mm LiCl, 1 mm EDTA, 0.1% SDS, 1% Triton X-100, and protease inhibitors), and once in a TE, 50 mm NaCl buffer. The immune complexes were harvested in elution buffer (10 mm Tris-HCl (pH 7.5), 1 mm EDTA, 1% SDS and 100 mm DTT). For re-ChIP assays, the eluted complexes were diluted 1:10 with dilution buffer (10 mm Tris-HCl (pH 7.5), 150 mm NaCl, 1 mm EDTA, 0.5% sodium deoxycholate, 1% Nonidet P-40, and protease inhibitors), and the steps of ChIP assay were repeated. Cross-linking was reversed by incubation overnight at 65 °C with a 0.04 volume of 5 m NaCl. After treatment with proteinase K (Wako), the DNA was purified using Nucleospin Extract II (Macherey-Nagel, Düren, Germany) and quantified by real time PCR on a 7300 Real Time PCR System (Applied Biosystems, Bedford, MA) using Premix Ex Taq (Probe qPCR) or SYBR Premix Ex Taq (Takara Bio, Shiga, Japan). The specific primer pairs and probes used in this study are listed in supplemental Table S1.

Other Procedures

Details of the materials and methods used in this study can be found in the supplemental experimental procedures.

RESULTS

High Expression of Each Member of the d4 Family and PHF10 Enhances the Transactivating Activity of RelA/p50, RelB/p52, and c-Rel/p50

We previously generated a 293FT cell line stably harboring an exogenous expression unit composed of tandem NF-κB-responsive elements, a minimum promoter (MinP), and the downstream luciferase (Luc) reporter gene (293FT-NF-κB-MinP-Luc) (Fig. 1A) (Ref 19), in which the NF-κB-binding sites were derived from HIV-1 LTR (Fig. 1B). Whereas this reporter cell system was useful in evaluating chromosome structure-dependent NF-κB transactivation, it was found not to be as sensitive in responding to the exogenous introduction of NF-κB. We thus isolated several additional cellular clones from this parental reporter cell population and tested each for Luc inducibility following TNF-α treatment. A clone, which we designated as 293FT-NF-κB-MinP-Luc-A3, (abbreviated as NF-κB-MinP-Luc-A3 hereafter), showed one of the highest levels of Luc inducibility by TNF-α and was selected for further analysis (supplemental Fig. S1).

FIGURE 1.

d4 family members are potentially involved in both the canonical and noncanonical NF-κB transactivation of SWI/SNF-dependent promoters. A and B, schematic representation of the constructs used in this study: NF-κB-MinP (A) and HIV-1 LTR (B). The binding sites for NF-κB and the TATA boxes are highlighted by black and white bars, respectively. Transcriptional start sites are designated as +1. Positions detected by real time PCR are indicated by black arrows and gray bars. The splicing donor and the acceptors are indicated by white and black circles. C, NF-κB-MinP-Luc-A3 cells were cotransfected with 10 ng of vectors expressing NF-κB subunits and 490 ng of candidate protein expressing or respective control vectors in different combinations. Luciferase activity was measured 48 h after transfection and normalization using control plasmid transfectants. D, NF-κB-MinP-Luc-A3 cells were transfected with 500 ng of vectors expressing short hairpin RNAs against each candidate protein and also Brm or BRG1. At 42 h post-transfection, cells were treated with 10 ng/ml TNF-α for 6 h, which led to a 1500-fold activation of luciferase activity compared with the control vector (CV-3) transfectants (data not shown). Luciferase activities were normalized to the CV-3 control transfectants. The corresponding target sites of the shRNAs against each of the candidates analyzed in this experiment are shown in supplemental Fig. S1. Lane 2, shDPF1-cds-1; lane 3, shDPF1-utr-1; lane 4, shREQ#1; lane 5, shREQ#2; lane 6, shDPF3-cds-1; lane 7, shDPF3a-utr-2; lane 8, shDPF3b-cds-3; lane 9, shPHF10-cds-1; and lane 10, shPHF10-utr-1. The results presented here are the average of at least three independent experiments, and the bars indicate the standard deviation.

Expression vectors for the five co-activator candidate proteins, four d4 family proteins and PHF10 (supplemental Fig. S2), as well as the empty vector (CV-1) as a control were transfected into NF-κB-MinP-Luc-A3 together with each pair of plasmids expressing the NF-κB dimer, RelA/p50, RelB/p52, or c-Rel/p50, as well as the control vector for NF-κB expression (CV-2). The expression of these five candidate proteins that were tagged with FLAG at their N termini were confirmed by immunoblotting (supplemental Fig. S3A). As shown in Fig. 1C, the basal Luc activity in cells transfected with CV-2 was not significantly affected by the introduction of any of the candidate proteins (lanes 2–6 compared with lane 1). Compared with the Luc activity of cells transfected with CV-1, the introduction of RelA/p50, RelB/p52, and c-Rel/p50 increased these reporter expression levels by 42-, 6-, and 3-fold, respectively (Fig. 1C, lanes 13, 19, and 7 compared with lane 1). These results indicated that NF-κB-MinP-Luc-A3 has very low basal NF-κB activity and that none of the candidates have any effects alone under these conditions, even if expressed at high levels. Importantly, transactivation mediated through RelA/p50 (Fig. 1C, lanes 14–18 compared with lane 13), RelB/p52 (lanes 20–24 compared with lane 19), and c-Rel/p50 (lanes 8–12 compared with lane 7) was enhanced by the endogenous expression of any of the five candidate proteins (lanes 7–24). No significant differences were found among these five proteins in terms of the enhancement of any of the three NF-κB dimers, suggesting that all potentially function as co-activators of NF-κB in any context, at least when expressed at high levels.

Both DPF3a and DPF3b Are Required for SWI/SNF-dependent Transcriptional Activation through the NF-κB Canonical Pathway

We concluded that transfection experiments involving NF-κB expression vectors would not fully reflect NF-κB activation induced by naturally occurring signal transduction pathways and that the transient expression of candidate proteins by plasmid vectors is very high and thus far from physiological (supplemental Fig. S4). Hence, the experimental conditions used above would most likely not reflect the functional specificity of these proteins correctly. We thus examined the requirement for each candidate protein in the activation of NF-κB by TNF-α treatment, which strongly and almost exclusively activates the endogenous RelA/p50 dimer. Because 293FT cells express all of the transcripts coding these five candidate proteins, as judged by semi-quantitative RT-PCR (supplemental Fig. S3B and data not shown), we performed knockdown experiments using the same assay cell line, NF-κB-MinP-Luc-A3. The efficiency and specificity of each of the designed shRNAs were confirmed by semi-quantitative RT-PCR or immunoblotting using 293FT cells (supplemental Fig. S3, B and C). The NF-κB-MinP-Luc-A3 cell line was transfected with a vector that expresses shRNA targeting each candidate protein and the catalytic subunits of the SWI/SNF complex as well as a control shRNA vector (CV-3). At 42 h after transfection, cultures were treated with or without TNF-α for an additional 6 h.

Upon TNF-α stimulation, Luc activity increased by 1400-fold when the cells were transfected with CV-3. The knockdown of Brm and BRG1, the alternative catalytic subunits of the SWI/SNF complex, caused a significant reduction in transactivation by 82 and 66%, respectively (Fig. 1D, lanes 11 and 12), indicating that NF-κB-MinP in the reporter cell line requires the SWI/SNF complex to be activated through the NF-κB canonical pathway. The depletion of DPF2 (Fig. 1D, lanes 4 and 5) and PHF-10 (Fig. 1D, lanes 9 and 10) lowered the ability of TNF-α to activate this promoter by 60 and 40% respectively, and the knockdown of DPF1 (Fig. 1D, lanes 2 and 3) marginally but clearly reduced the Luc promoter activity. These results are consistent with the notion deriving from the results of the overexpression experiments (Fig. 1C) that DPF1, DPF2, and PHF10 contribute to transactivation via the endogenous RelA/p50 dimer and the difference among the effects of each knockdown might be somehow reflecting their endogenous levels. Interestingly, the simultaneous depletion of DPF3a and DPF3b by a single shRNA, whose targeting site is within the shared region for these proteins, was shown to reduce TNF-α-dependent activation by 84% (Fig. 1D, lane 6). Furthermore, a single knockdown of either DPF3a or DPF3b also caused a comparative reduction in promoter activation (Fig. 1D, lanes 7 and 8). These results suggest that in these cells, among the tested candidates, either DPF3a or DPF3b have the most crucial role in the canonical NF-κB pathway, which cannot be efficiently compensated for by the other candidate proteins. The overall results shown in Fig. 1D were almost identically obtained using another 293FT clone K5 (supplemental Fig. S5). Taken together, these data support the idea that in 293FT cells DPF3a and DPF3b are the most critical factors required for RelA/p50 to activate this NF-κB-containing promoters in an SWI/SNF-dependent manner, and we therefore decided to further concentrate our analysis on DPF3a and DPF3b.

Both DPF3a and DPF3b Directly Bind to RelA and p50 as Well as Several Subunits of the SWI/SNF Complex in Vitro

Because DPF3a and DPF3b were found to be required for SWI/SNF complex-dependent RelA/p50 transactivation, we next evaluated the proteins with which DPF3a and DPF3b can associate in vitro. We first performed glutathione S-transferase pulldown assays using purified GST alone, GST-DPF3a, and -DPF3b and several subunits of the SWI/SNF complex translated in vitro in the presence of [³⁵S]methionine (supplemental Fig. S6). Both DPF3a and DPF3b directly bound to Brm, BRG1, Ini1, and BAF60a but not to β-actin, which are similar binding properties to those of DPF2 that we have previously reported (19).

We next performed a similar assay using in vitro translated RelA and p50 and GST-fused DPF3a and DPF3b as well as DPF2. As shown in Fig. 2A, either RelA or p50 alone as well as in combination with both proteins synthesized in the same translating mixture were found to directly associate with DPF3a and DPF3b. We also found that RelA or p50 alone also binds to DPF2 to a lesser extent. These binding properties of DPF2 are fully consistent with our previous report (19). In the current experiments, these two GST proteins were tagged with His₆ at their C termini to enable higher purification and the performance of a more sensitive binding assay (supplemental Fig. S7). We performed further GST pulldown assays using the same purified GST proteins and 293FT whole cell extracts obtained from cells stimulated with or without TNF-α as the input proteins. The extracts were treated with benzonase endonuclease to avoid detecting interactions mediated through DNA or RNA fragments. The pulled down cellular proteins were analyzed by immunoblotting, and as shown in Fig. 2B, both DPF3a and DPF3b were found to associate with endogenous RelA and BRG1. Importantly, the binding affinities for DPF3a and DPF3b in these cases were not significantly increased by TNF-α stimulation. It should be pointed out here that in the total protein preparations subcellular localization of natural NF-κB is disrupted. These results therefore suggest that the binding potential of RelA to DPF3a and DPF3b is not significantly enhanced by post-translational modifications of RelA triggered by TNF-α stimulation (25–28). Taken together, our findings indicate that both DPF3a and DPF3b have the potential to directly associate with RelA/p50 and the SWI/SNF complex and further that the binding potential is not enhanced by post-translational modifications of NF-κB or the SWI/SNF complex triggered by TNF-α.

FIGURE 2.

DPF3a and -3b associate with both RelA/p50 and the SWI/SNF complex. A, [³⁵S]methionine-labeled RelA and p50 were translated using a wheat germ extract system in vitro and incubated with GST alone and GST-His-tagged DPF2, DPF3a, or DPF3b. These interactions were analyzed by SDS-PAGE followed by autoradiography. B, 293FT cells were stimulated with or without 10 ng/ml TNF-α for 10 min, and whole cell lysates were then obtained and subjected to benzonase endonuclease treatment. The lysate was incubated with GST alone, GST-His-tagged DPF2, DPF3a, or DPF3b, and pulled down materials were analyzed by immunoblotting. C, 293FT cells were stimulated with or without 10 ng/ml TNF-α for 10 min, and whole cell lysates were obtained and treated with endonuclease. Immunoprecipitation (IP) assays were performed using α-FLAG antibodies, and precipitated materials were eluted with free FLAG peptide and analyzed by immunoblotting (IB). D, 293FT cells stably expressing FLAG, FLAG-DPF3a, or FLAG-DPF3b were stimulated with or without 10 ng/ml TNF-α for 60 min. Nuclear extracts were then prepared and subjected to immunoprecipitation assays with α-FLAG antibodies followed by immunoblotting. E, 293FT cells were treated with 10 ng/ml TNF-α for 20 min, and whole cell lysate was obtained and treated with endonuclease. Immunoprecipitation assays were performed using α-DPF3 antibody, and precipitated materials were analyzed by immunoblotting.

DPF3 Interacts with the SWI/SNF Complexes and RelA/p50 in the Nucleus

We prepared 293FT cells stably expressing either FLAG-tagged DPF3a or DPF3b through the introduction of retrovirus vectors (293FT-FLAG-DPF3a and 293FT-FLAG-DPF3b) and isolated nuclear fractions from each cell type with or without TNF-α treatment. We then performed a co-immunoprecipitation assay using α-FLAG antibodies (Fig. 2C). The subunits of the SWI/SNF complex, Brm, BRG1, and BAF155, were found to be co-immunoprecipitated with either DPF3a or DPF3b, independently of TNF-α stimulation. In the immunoprecipitates with FLAG-DPF3b, both RelA and p50 were detected specifically in TNF-α-stimulated cells, indicating that DPF3b efficiently associates with both RelA and p50, which are recruited to the nucleus. In the immunoprecipitates with FLAG-DPF3a from TNF-α-stimulated cells, p50 was detected, whereas only marginal levels of RelA were detectable, suggesting that DPF3a binds RelA/p50 at a lower affinity than DPF3b in the nucleus. Our observations therefore suggest that the association between DPF3a/b and RelA/p50 occurs following RelA/p50 translocation to the nucleus where DPF3a or -3b consistently interacts with the SWI/SNF complex. When similar analyses were performed on the other candidate proteins, DPF1 and DPF2 showed similar binding properties to DPF3a and -3b for the associations with the SWI/SNF complex and RelA/p50 (supplemental Fig. S8).

To further analyze the DPF3 interaction with the RelA/p50 and SWI/SNF complex, we prepared whole cell extracts from 293FT-FLAG-DPF3a and 293FT-FLAG-DPF3b with or without TNF-α treatment (Fig. 2D). Each cellular extract was again treated with benzonase endonuclease and then subjected to a co-immunoprecipitation assay using α-FLAG antibodies. Brm, BRG1, BAF155, and BAF60a were all successfully co-immunoprecipitated from the total cellular lysates of both 293FT-FLAG-DPF3a and -DPF3b independently of TNF-α. RelA was found to be co-immunoprecipitated with DPF3b and to a lesser extent with DPF3a. DPF3a and DPF3b would associate with RelA after the preparation of the whole cell extracts from unstimulated cells, where the natural NF-κB subcellular localizations are disrupted. These data again support that the RelA binding potential to DPF3a or DPF3b was not significantly affected by TNF-α treatment, consistent with the results shown in Fig. 2B.

To finally show our observations that DPF3a/b associates with RelA/p50 or the SWI/SNF complex is valid even in nonmanipulated cells, we treated 293FT cells with TNF-α for 20 min and immunoprecipitated endogenous DPF3a and DPF3b with antibody against DPF3, which does not discriminate between DPF3a and DPF3b. The immunoprecipitates were analyzed, and endogenous DPF3a/b was shown to associate with RelA, BRG1, and BAF155, just like exogenously introduced DPF3 proteins (Fig. 2E).

Kinetics of RelA Recruitment upon TNF-α Stimulation Correlate with Those of Primary Transcript Production from Both Artificial and Natural Promoters

To understand molecular mechanisms involved in transcriptional activation after TNF-α stimulation, we next analyzed the kinetics of RelA recruitment and transactivation. Following cellular exposure to TNF-α, it was previously reported at a single cell resolution that RelA oscillates dynamically between the nucleus and the cytoplasm and that this oscillation cycle is about 1 h (29). To analyze the primary effects of TNF-α stimulation on RelA/p50 transactivation, we performed ChIP analysis of the promoters of the reporter genes after stimulation within 1 h. In NF-κB-MinP-Luc-A3 cells, the results showed that the stimulation-dependent recruitment of RelA achieved a peak at 30 min post-treatment and drastically reduced to marginal levels within 60 min (Fig. 3A). Next, by using 293FT cells harboring an HIV-1-based reporter provirus (293FT-LTR-Luc-5; Fig. 1B) instead, we performed the same set of analyses (Fig. 3B). Interestingly, RelA recruitment to the NF-κB-responsive elements within the HIV-1 LTR kept increasing during the first 60 min after TNF-α stimulation with small leakage of the signal detected in the nucleosome 1 (nuc-1) region. Considering that the nucleotide sequences of the NF-κB elements are identical in both of these promoters, it is noteworthy that distinct kinetics of RelA recruitment were observed, i.e. the wild type HIV-1 LTR retains RelA longer.

FIGURE 3.

Kinetics of RelA recruitment upon TNF-α stimulation correlate with the primary transcript production levels from both artificial and native promoters. A, B, and E, kinetics of RelA recruitment to NF-κB-MinP (A), HIV-1 LTR (B), and the endogenous IL-6 promoter (E) upon TNF-α stimulation. Cells were treated with 10 ng/ml TNF-α for the indicated times and were harvested and precipitated by antibodies against RelA or with a normal rabbit IgG control. The immunoprecipitated DNA was quantified by real time PCR and normalized relative to the input. C and F, real time RT-PCR analysis of transcripts of HIV-1 LTR (C) and the endogenous IL-6 promoter (F). Cells were treated with 10 ng/ml TNF-α for the indicated times, and mRNA was prepared. Spliced and unspliced mRNAs were quantified, and these levels were normalized to individual untreated samples. D, schematic representation of the IL-6 promoter region. The binding site for NF-κB and other transcription factors and the TATA box are highlighted by black, gray, and white bars, respectively. Transcriptional start sites are designated as +1. Positions detected by real time PCR are indicated by black arrows and a gray bar.

We next analyzed whether there was any association between the recruitment levels of RelA and the transactivation kinetics of these promoters. By quantifying the HIV-1 LTR transcript levels at the indicated time points by real time RT-PCR, we found that the kinetics of the de novo synthesized primary unspliced mRNAs were very similar to those of RelA recruitment, whereas those of the mature spliced mRNAs kept increasing (Fig. 3C). Because the primary transcript of NF-κB-MinP does not contain introns, it cannot be distinguished from the mature mRNA. Interestingly, using a promoter of an endogenous TNF-α-inducible gene, IL-6 (Fig. 3D), which requires the SWI/SNF complex for activation (14), we observed a good correlation in kinetics between RelA recruitment and primary transcript synthesis after TNF-α stimulation (Fig. 3, E and F, and the induction kinetics is more prompt than those observed for HIV-1 LTR in the same cells (Fig. 3B). In addition, the kinetics of the recruitment of RelA to the IL-6 promoter are similar to those of NF-κB-MinP, achieving a peak at 30 min after stimulation and decreasing at 60 min. Our present observations thus suggest that HIV-1 LTR retains RelA longer against the oscillation cycle of RelA between the nucleus and the cytoplasm. HIV-1 LTR contains many binding sites for other transcription factors, some of which would associate with RelA (30–32). This might cause differences in the kinetics of the artificial MinP, which is simply under the control of NF-κB. However, the endogenous IL-6 promoter is also regulated by other transcription regulators but showed no retention of RelA. This observation might be reflecting that, unlike HIV-1 LTR, this gene is expected to be strictly down-regulated after the stimulation for proper inflammatory response.

Chromatin Dynamics around the NF-κB-binding Sites in Both Artificial and Native Promoters after TNF-α Treatment

We next analyzed the dynamics of endogenous DPF3 and the SWI/SNF complex as well as RelA/p50 for both NF-κB-MinP and HIV-1 LTR in response to TNF-α stimuli. Consistent with Fig. 3A, this stimulation caused a dynamic recruitment of RelA to NF-κB-MinP at 30 min (left panel of Fig. 4A). Similar amounts of NF-κB p50 subunit were detected both before and after TNF-α stimulation, which we think is consistent with previous observations that low levels of the p50 subunit are present on specific NF-κB-responsive promoters in the nucleus of unstimulated cells as the p50 homodimer, whereas the majority of this subunit is retained in the cytoplasm as RelA/p50 (33–35). The p50 homodimer itself does not contain transactivation domains, but it associates with promoter regions of some NF-κB-dependent genes, including integrated HIV-1 LTR in the unstimulated conditions (36). However, Brm, BRG1, and DPF3 were all detected at the promoter even before stimulation, and their recruitment levels were not significantly changed upon exposure to TNF-α, apart from Brm, which showed elevated levels at the promoter (right panel of Fig. 4A). In the case of 293FT-LTR-Luc-5 cells, RelA was recruited to NF-κB elements in the HIV-1 LTR within 10 min of stimulation, whereas the NF-κB p50 subunit was found to be constitutively present in the LTR (upper panel of Fig. 4B). Brm, BRG1, and DPF3 were also detected at the promoter prior to TNF-α treatment, and their recruitment levels were reduced by about 50% after stimulation, suggesting that upon transcriptional activation, the SWI/SNF complex partially releases from the promoter with DPF3 (lower panel of Fig. 4B). Because other subunits of the complex such as BAF60a and Ini1 behaved in a similar manner to Brm or BRG1 (supplemental Fig. S9), we speculate that the changes in the levels of the catalytic subunits associated with the HIV-LTR promoter were as part of the SWI/SNF complex.

FIGURE 4.

Chromatin dynamics and recruitment of DPF3 and RelA/p50 to SWI/SNF-dependent promoters upon TNF-α treatment. ChIP analysis was performed, and the data were quantified as described in Fig. 3. A, ChIP analysis of RelA, p50, Brm, BRG1, and DPF3 levels at the NF-κB-MinP upon TNF-α stimulation. B, recruitment of RelA, p50, Brm, BRG1, BAF60a, Ini1, and DPF3 to the regions within the HIV-1 LTR and GAPDH promoter after TNF-α treatment. Results presented here are the average of at least two independent experiments. ND, not detectable.

Whereas the two promoters under analysis showed different RelA recruitment kinetics, they were found to share two important features from our observations. First, endogenous DPF3 together with the SWI/SNF complex is bound to the promoters prior to TNF-α stimulation. Second, upon stimulation, RelA/p50, which can directly interact with DPF3 (Fig. 2A), was found to be rapidly recruited to the promoters, and this recruitment was shown to be associated with transcriptional activation, as judged by the primary transcript levels.

Finally, we analyzed the endogenous IL-6 promoter to check whether DPF3 links the SWI/SNF complex and RelA/p50 forming a larger complex. ChIP analysis showed that both RelA and DPF3a/b were recruited to the promoter in a TNF-α-dependent and -independent manner, respectively (Fig. 5A). To perform sequential ChIP analysis, the immunoprecipitates with α-RelA antibody were re-immunoprecipitated with antibodies specific for DPF3a/b and Brm as well as the α-RelA antibody (positive control) and normal rabbit IgG (negative control), respectively (Fig. 5B). The second round of immunoprecipitates using α-DPF3 and Brm antibodies indicated that DPF3 and the SWI/SNF complex co-associate with RelA on the IL-6 promoter after the RelA/p50 heterodimer is recruited upon TNF-α stimulation. In conclusion, all these results support that DPF3a and -3b link the SWI/SNF complex to NF-κB to transactivate the target promoters that require chromatin remodeling for the transcriptional initiation.

FIGURE 5.

ChIP and re-ChIP assays at the endogenous IL-6 promoter. A, ChIP analysis of RelA and DPF3 occupancy levels at the IL-6 promoter. 293FT cells were treated with or without 10 ng/ml TNF-α for 30 min and harvested for precipitation by antibodies against RelA and DPF3. B, primary immunoprecipitated materials concentrated with α-RelA antibody were followed by secondary rounds of immunoprecipitations with indicated antibodies. Immunoprecipitated materials were quantified by real time PCR.

DISCUSSION

By extending our previous observation that DPF2 links RelB/p52 and the SWI/SNF complex for NF-κB transactivation via its noncanonical pathway, we have here shown that each of the five proteins, DPF1, DPF2, DPF3a, DPF3b, and PHF10, can equally function as an efficient co-activator of the RelA/p50 dimer at the most downstream part of the NF-κB canonical pathway, when they are exogenously expressed at high levels (Fig. 1C). Under these conditions, each of these proteins activated the other two representative NF-κB dimers, RelB/p52 and c-Rel/p50, with similar efficiencies. When these five proteins were expressed by retrovirus vectors, they are constitutively binding to the SWI/SNF complex in the nucleus and begin to associate with RelA/p50 upon TNF-α treatment (Fig. 2C and supplemental Fig. S8). From these results, we think each of these proteins would function as a co-activator of NF-κB in certain tissues. DPF1 was reported to be neurospecific and play an important role in developing neurons (20, 24). DPF2, which is ubiquitously expressed, is known to be involved in the processes of apoptosis in mouse myeloid cells following interleukin 3 (IL-3) deprivation (37). In normal cells, DPF3a/b was reported to be expressed specifically in cardiac and skeletal muscle and was critical for heart and muscle development (23). Because NF-κB plays important roles in programmed cell death and development, these d4 proteins might be involved in these physiological phenomena through activating some specific NF-κB dimers. Because DPF1, DPF2, DPF3b and PHF10 contain double PHD fingers, which are not possessed by any of the core subunits of the SWI/SNF complex, they are likely to play some roles in selecting specific species of modified histones in the nucleosomes at target promoter regions.

In 293FT cells, which express only low levels of endogenous mRNAs for the five candidate protein co-activators, the knockdown of either DPF3a or DPF3b suppressed TNF-α-induced NF-κB transactivation to marginal levels, suggesting that both proteins are essential for the NF-κB canonical pathway (Fig. 1D). Currently, we cannot fully explain this observation, but there could be several possibilities. One possibility would be that the SWI/SNF complex needs to bind both DPF3a and DPF3b to perform its full function for NF-κB transactivation. We would, however, prefer another possibility as follows. DPF3a and DPF3b share roles in NF-κB transactivation probably through their common N-terminal regions. The endogenous cellular amounts of both proteins in 293FT cells are so low (supplemental Fig. S3) that depletion of either would drastically reduce their recruitment frequency to the promoter of the reporter, where a single molecule of either DPF3a or DPF3b is sufficient for the SWI/SNF complex to perform full transactivation. Because DPF3a lacks functional PHD fingers, however, it is also possible that they would show distinct promoter preferences in some cases, as has been shown previously by ChIP analysis of exogenously expressed DPF3a and DPF3b (23).

We have previously shown that a representative target gene of the NF-κB canonical pathway, IL-8, is induced in HeLa cells by TNF-α and that this induction is not affected by the knockdown of DPF2. We suggested from this that DPF2 does not significantly contribute to the NF-κB canonical pathway, at least in this cell system. Because the endogenous IL-8 gene was found not to be affected by the knockdown of either catalytic subunit of the SWI/SNF complex (19), we believe that this promoter is induced independently of SWI/SNF. In immune cells, some NF-κB target promoters have also been reported not to require chromatin remodeling for their transcriptional initiation upon LPS stimulation, because they already possess an open chromosome structure (38).

Importantly, with respect to our current investigations, the results of ChIP analysis using antiserum that we prepared indicated that the endogenous DPF3 proteins (this antiserum does not discriminate between DPF3a and DPF3b) as well as the SWI/SNF complex are recruited continuously to the promoters we examined, NF-κB-MinP and wild type HIV-1 LTR. Considering that both DPF3a and DPF3b directly bind to p50 in vitro, we speculate that DPF3a/3b together with the SWI/SNF complex locate near the NF-κB-binding sites through the binding of the p50 homodimer under unstimulated conditions but through active RelA/p50 upon stimulation. This dimer substitution would then trigger SWI/SNF-dependent transactivation. Hence, the SWI/SNF complex would be present at these target promoters to facilitate the ready recruitment of RelA/p50 for the prompt remodeling of the surrounding chromatin structures. Considering our observations that transcriptional enhancement of HIV-1 LTR occurs just after the SWI/SNF complex and RelA/p50 get together on the promoter (Figs. 3, B and C, and 4B), DPF3a/3b is significant for this activation as a linker protein.