Diverse Effects of BCL3 Phosphorylation on Its Modulation of NF-κB p52 Homodimer Binding to DNA*

IκB proteins control the subcellular localization and DNA binding activity of NF-κB transcription factors. BCL3 is a nuclear IκB that can inhibit or enhance the binding of NF-κB p50 or p52 homodimers to consensus DNA-binding (κB) sequences or form a κB-binding complex with homodimers. To study BCL3 function, we have used gel shift analysis and tagged protein and tagged DNA coprecipitation analyses. Our results show that at intermediate ratios of BCL3 to p52 all observed phosphoforms of BCL3 are able to form a κB-binding complex with p52 homodimers. At low BCL3/p52 ratios, BCL3 increases the rate of p52 homodimer binding to κB sites in the presence of nonconsensus DNA and dissociates from the complex. At high BCL3/p52 ratios, BCL3 forms a higher order inhibitory complex with p52 homodimers. All of these effects depend on BCL3 phosphorylation and relative concentration. These results indicate that BCL3 phosphorylation may affect its regulation of NF-κB-dependent transcription in vivo.

IB proteins control the subcellular localization and DNA binding activity of NF-B transcription factors. BCL3 is a nuclear IB that can inhibit or enhance the binding of NF-B p50 or p52 homodimers to consensus DNA-binding (B) sequences or form a B-binding complex with homodimers. To study BCL3 function, we have used gel shift analysis and tagged protein and tagged DNA coprecipitation analyses. Our results show that at intermediate ratios of BCL3 to p52 all observed phosphoforms of BCL3 are able to form a B-binding complex with p52 homodimers. At low BCL3/p52 ratios, BCL3 increases the rate of p52 homodimer binding to B sites in the presence of nonconsensus DNA and dissociates from the complex. At high BCL3/p52 ratios, BCL3 forms a higher order inhibitory complex with p52 homodimers. All of these effects depend on BCL3 phosphorylation and relative concentration. These results indicate that BCL3 phosphorylation may affect its regulation of NF-B-dependent transcription in vivo.
BCL3 is a member of the IB family of inhibitors of NF-B transcription factors (1)(2)(3). IB proteins control NF-B activity in at least two ways: IB can anchor NF-B in the cytoplasm of resting cells by blocking its nuclear localization signal and can inhibit NF-B binding to consensus DNA-binding (B) sites. However, under some circumstances IB proteins can associate with NF-B without blocking the nuclear localization signal (4 -6); they can form an IB-NF-B-B complex (6 -8), and they can enhance NF-B binding to B sites (9 -11). BCL3 has been demonstrated to have each of these "nonclassical" IB activities, but the mechanisms of BCL3 action have not been elucidated.
BCL3 is unique among IB proteins in that it is a nuclear protein (4,5,7) that specifically associates with homodimers of p50 or p52 subunits (4, 28 -32) and that contains N-and Cterminal regions that can act as transactivation domains (7). p50 and p52 are similar in their primary structures, and they are unique among NF-B family members in that they are processed from precursors (NF-B1 p105 and NF-B2 p100) with C-terminal IB regions, have no defined transactivation domains, and are found in homodimeric form in both the cytosol and the nucleus of resting cells. These homodimers may function to competitively inhibit B binding by transactivating NF-B dimers. BCL3 can 1) remove homodimers from B sites so that transactivating NF-B dimers can bind and act as an antirepressor (28,33); 2) form a complex with homodimers at B sites and act as a transactivator (7,8); or 3) enhance homodimer binding to B sites (11). Thus, BCL3 may repress transcription indirectly by increasing B-site occupancy by p50 or p52 homodimers or it can directly or indirectly activate transcription. Whether or not BCL3 activity is regulated in the cell is not known.
In this work, we have studied the mechanisms by which BCL3 modulates DNA binding by p52 homodimers. We have asked the following questions: 1) which regions of BCL3 are phosphorylated; 2) how does BCL3 phosphorylation affect its ability to form a BCL3-(p52) 2 -B complex; 3) which region(s) of BCL3 are necessary to inhibit p52 homodimer binding to DNA;4) what is the mechanism of BCL3 inhibition of homodimer binding to DNA; and 5) what is the mechanism of BCL3 enhancement of the formation of (p52) 2 -B complexes? Our results show that BCL3 can have a variety of effects on p52 homodimer binding to DNA, depending on BCL3 concentration and phosphorylation. The results are consistent with a model in which BCL3 has a versatile role in the regulation of NF-Bdependent gene expression.

MATERIALS AND METHODS
Plasmids, Antibodies, and B Probe-H 6 BCL3 and H 6 BCL3⌬N plasmids have been described previously (5), and the plasmids pMT2Tp52, pMT2TBCL3 (wtBCL3), and pMT2TBCL3⌬C were a gift from U. Siebenlist (7). Antibodies to p52 (I-18) or to the C terminus of BCL3 were purchased from Santa Cruz. Antiserum to the N terminus of BCL3 has been described previously (5); rabbit polyclonal BCL3 ARD antiserum was made by immunizing rabbits with a glutathione S-transferase-ARD fusion protein (Josman Laboratories, Napa, CA). The B probe used for all studies was prepared by annealing a 27-mer B oligonucleotide (5Ј-CAACGGCAGGGGAATTCCCCTCTCCTT-3Ј) to a 5Ј-biotinylated 21-mer primer oligonucleotide (5Ј-biotin-AAGGAGAGGG-GAATTCCCCTG-3Ј) and filling with dNTPs and Klenow enzyme. For labeled probe, [␣-32 P]dCTP was substituted for dCTP.
Cell Culture, Transfections, and Metabolic Labeling-COS7 cells (ATCC) were maintained as described (5) and transfected using a Eurogentec Cellject electroporator with pulse values set at 300 V, 960 microfarads, and infinite resistance. The transfection buffer consisted of protein-free Dulbecco's modified Eagle's medium (Sigma), 5-10 g of plasmid, and 4 g/ml DEAE-dextran (34). Typically, 5 ϫ 10 6 cells were electroporated in 800 l of media. For 32 P labeling of H 6 BCL3, transfected cells were lifted 2 days after transfection, washed twice in phosphate-buffered saline, washed once in serum-free, phosphate-free Dulbecco's modified Eagle's medium (Sigma), and then incubated in 1 ml of this medium for 10 min prior to the addition of 0.3-0.6 mCi/ml [ 32 P]orthophosphate (NEN Life Science Products). Cells were incubated in this medium for 2-3 h, washed twice in medium, and lysed in Nonidet P-40 lysis buffer as described below. 32 P-H 6 BCL3 was purified with nickel-nitrilotriacetic acid (NTA) agarose beads as described below.
Graded Dephosphorylation of BCL3-Typically, 1 unit of calf intestinal phosphatase (CIP) (Sigma) per mg of total protein in control or BCL3-transfected cellular extract was incubated in CIP reaction buffer (60 mM NaCl, 2 mM MgCl 2 , 2 mM DTT, 20 mM Hepes, pH 8.0, protease inhibitors, 10 mM EDTA, 10 mM NaF) for 1 h at room temperature. EDTA and NaF were added at different time points to inhibit CIP activity. For mock-treated BCL3 samples, CIP was boiled or incubated in EDTA for 20 min prior to use.
Partial Purification of H 6 BCL3-Nickel-NTA beads (Qiagen) were preequilibrated in buffer L plus 1% bovine serum albumin and 0.1 mM imidazole. Typically, 1 ml of H 6 BCL3 in whole cell extracts and 12 ml of buffer L were incubated with 100 l of nickel-NTA beads on ice for 2 h with gentle rocking. Beads were pelleted and then washed 6 times (8 min each) with 5 ml of buffer L supplemented with 4 mM imidazole before H 6 BCL3 was eluted in 400 l of buffer L with 90 mM imidazole.
Protease Digestions-For exhaustive protease digestions, total protein in BCL3-transfected cell extracts was diluted to 1 g/l in buffer L and then boiled in 1% SDS and 5 mM DTT (purified BCL3 was not diluted). After cooling to room temperature, 5 g of LysC (Sigma) or GluC (Sigma) per g of substrate protein was added. Digestion reactions were allowed to incubate at 37°C for 2 h before the addition of 0.1 ϫ volume of 4 ϫ sample loading buffer, boiling, separation by SDS-PAGE, electrotransfer to a PVDF membrane, autoradiography of 32 P-labeled peptides, and Western blotting. For protease digestion of native BCL3, purified BCL3 was incubated in digestion buffer (70 mM Tris, pH 7.5, 100 mM NaCl, 5 mM DTT) and titrated with LysC protease (0.5-5 g of LysC per g of purified BCL3). Digestion reactions were stopped after 1 h by boiling in sample loading buffer, and peptides were analyzed by SDS-PAGE and Western blotting.
Gel Shift Assays-These were performed as described previously (5), except that sonicated placental DNA or salmon testes DNA was used as nontarget competitor DNA to reduce probe binding by nonspecific DNAbinding proteins, the final concentration of NaCl was 60 mM, and EDTA was added to the binding buffer to 4 mM when CIP-treated or mocktreated BCL3 was added to the binding reaction. For a standard gel shift assay, proteins were preincubated with competitor DNA in binding buffer (60 mM NaCl, 1-4 mM EDTA, 5% glycerol, 1 mM DTT, 40 mM Tris, pH 7.4) for 30 min before the addition of an excess (1 pmol/sample) of 32 P-labeled B probe and then allowed to equilibrate for 40 min before electrophoresis was performed in a 4.5% nondenaturing poly-acrylamide gel in 25 mM Tris, 190 mM glycine, 1 mM EDTA.
Precipitation of B-bound Complexes with Biotinylated B Probe-All reactions were performed at room temperature. 0.8 pmol of 5Јbiotinylated B probe per sample was incubated and rocked in 100 l of binding buffer (60 mM NaCl, 1 mM DTT, 5 mM EDTA, 10% glycerol, 0.01 mg/ml sonicated placental DNA, 2 mg/ml bovine serum albumin, 40 mM Tris, pH 7.4) for 30 min with 2 l of avidin-coated Dynal beads. Beads were pelleted and washed once in 100 l of binding buffer and then incubated with p52-transfected or untransfected nuclear extracts in binding buffer for 30 min. Beads were pelleted, resuspended in binding buffer, and aliquoted into sample tubes before the addition of BCL3transfected or untransfected whole cell extracts. The BCL3 binding reaction was allowed to proceed for 20 min; proteins were eluted from pelleted beads by boiling in sample loading buffer and analyzed by Western blotting.
Sandwich Coprecipitation Assays-These were performed at room temperature. Typically, 300 l of H 6 BCL3 in whole cell extracts was immobilized on an equal volume of nickel-NTA beads, washed once with CIP buffer (in the absence of NaF and EDTA), and incubated in CIP buffer with CIP for 45 min. Beads were pelleted and washed extensively in buffer L supplemented with 10 mM NaF, and then washed twice in binding buffer (40 mM Hepes, pH 7.5, 60 mM NaCl, 5% glycerol, 1% bovine serum albumin, 1 mM DTT) before the addition of p52-transfected or untransfected nuclear extracts in binding buffer. After 40 -60 min, the solution was aliquoted into separate sample tubes; beads were pelleted at 5,000 rpm for 2 min. Then wtBCL3-transfected and untransfected whole cell extract, diluted to 60 mM NaCl in binding buffer, was added to a final volume of 200 l and allowed to incubate and rock gently for 3 h. Finally, beads were pelleted; precipitated proteins were eluted with elution buffer (200 mM NaCl, 40 mM Hepes, pH 8.0, 1 mM DTT); and samples were prepared for SDS-PAGE by boiling in sample buffer.

RESULTS
The C-terminal Domain of BCL3 Is Extensively Phosphorylated-BCL3, with a predicted molecular mass of 48 kDa, migrates in SDS-PAGE as a major band with an apparent molecular mass of approximately 60 kDa and two minor bands of approximately 55 and 65 kDa, when derived from transiently transfected COS7 cells (Fig. 1A). COS7 cell-derived BCL3 comigrates with BCL3 derived from a variety of cell lines (not shown). Graded dephosphorylation of BCL3 with CIP, as described under "Materials and Methods," results in a shift of the 65-and 60-kDa forms to more rapidly migrating forms, indicating that BCL3 is constitutively phosphorylated, consistent with previous results (8,11,32). Only the 60-kDa phosphoform of BCL3 is detectable in most preparations, as shown in lane 4, and only preparations with low levels of 55-and 65-kDa forms were used for functional studies, unless otherwise indicated. BCL3 subjected to graded dephosphorylation has several intermediate phosphoforms, which indicates that BCL3 is extensively phosphorylated. We will refer to the phosphorylated 60-kDa form of BCL3 as "BCL3 P " and to the CIP-dephosphorylated 55-kDa form of BCL3 as "BCL3 D ." The BCL3 protein is composed of 16% (72 of 446 amino acids) serine, threonine, or tyrosine residues; 11 are in the N-terminal domain, 33 in the ARD, and 22 in the C-terminal domain (12). Serine alone represents 26% of C-terminal domain amino acids.
To determine which regions of BCL3 are phosphorylated, we used one-dimensional peptide mapping of proteolytically cleaved proteins. The BCL3 C-terminal peptide product of LysC protease digestion, Lys 353 -Ser 446 , is predicted to have a molecular mass of approximately 10 kDa (for the unphosphorylated peptide); this peptide spans the C-terminal domain and includes 6 amino acids of the 7th ankyrin repeat. The C-terminal peptide product of GluC digestion, Glu 277 -Ser 446 , which includes the C-terminal domain and ankyrin repeats 5, 6, and 7, is predicted to have a molecular mass of approximately 18 kDa. Glutamic acid residue Glu 336 was not vulnerable to GluC digestion in any of our studies.
BCL3 P in whole cell extracts was subjected to graded CIP dephosphorylation and exhaustive digestion with proteases. Fig. 1B shows a Western blot of C-terminal peptides of these samples. Dephosphorylated C-terminal peptides are not easily detected; they may be poorly retained by PVDF membranes.
The results show that the intermediate phosphoforms of BCL3 C-terminal peptides have distinctive mobilities, and the effect of dephosphorylation on peptide migration is similar to that seen with full-length BCL3 (compare Fig. 1, A and B). We conclude that the C-terminal domain of BCL3, like the Cterminal domains of IB␣ (10,18) and IB␤ (35), is extensively and constitutively phosphorylated.
To determine whether or not all of BCL3 phosphorylation is at the C-terminal domain, 32 P-labeled, nickel-NTA bead-purified BCL3 was subjected to graded LysC digestion. In native form, BCL3 P is highly resistant to LysC and GluC proteases, and complete digestion was not observed. The major labeled peptides detected by autoradiography (Fig. 1C, center panel) comigrate with BCL3 C-terminal peptides detected by Western blotting (left panel). No labeled bands are detectable which comigrate with ARD peptides (right panel).
To ensure complete proteolysis of BCL3, 32 P-labeled purified BCL3 samples were boiled in 1% SDS prior to being subjected to exhaustive protease digestion. In Fig. 1D, the left panel shows a Western blot with undigested (lanes 1 and 4), LysCdigested (lanes 2 and 5), or GluC-digested (lane 3) BCL3 proteins, blotted with antibody to the BCL3 C terminus. 32 P-Labeled peptides were detected by autoradiography of the same membrane, shown in the right panel. A sample containing predominantly BCL3 P is shown in lane 1; there is a low but detectable amount of BCL3 D . A histidine-tagged BCL3 construct lacking the N-terminal domain (H 6 BCL3⌬N) is shown in lane 4. The C-terminal peptide seen in lane 5 is the unphosphorylated H 6 BCL3⌬N C-terminal peptide; H 6 BCL3⌬N is poorly phosphorylated, and phosphorylated peptides are not detectable in reproductions of this Western blot, although they are detectable by autoradiography of 32 P-labeled peptides.
The results show that the major 32 P-labeled peptides visualized by autoradiography comigrate with C-terminal peptides detected by Western blotting, for both full-length and N-terminally truncated BCL3. In some experiments but not others, faint 32 P-labeled peptides are detectable that do not comigrate with C-terminal peptides. Their source is unknown. We conclude that most of BCL3 phosphorylation is at the C-terminal domain.
BCL3 Activity Is Phosphorylation-and Concentrationdependent-Previous results have indicated that BCL3 inhibition of p50 or p52 homodimer binding to B sites may be phosphorylation-and concentration-dependent. In gel shift assays, mammalian cell-derived BCL3 can form a BCL3-homodimer-B complex (7,8), but bacterially derived BCL3 inhibits p50 homodimer binding to B sites (4,5,29), which suggests that BCL3 phosphorylation may affect the formation of a BCL3-homodimer-B complex.
To study the concentration and phosphorylation dependence of the formation of a BCL3-(p52) 2 -B complex, we performed gel shift assays in which p52 was titrated with different phosphoforms of BCL3. Antiserum to the N terminus of BCL3 was added to decrease the mobility of BCL3-(p52) 2 -B complexes.
A typical result is shown in Fig. 2A. BCL3 P apparently has no ability to inhibit p52 homodimer binding to probe, even at high ratios of BCL3 P /p52 (lane 11), whereas BCL3 D is an efficient inhibitor (lane 13). At low BCL3/p52 ratios, p52 homodimer binding to probe is enhanced in the presence of BCL3 P (lanes 2 and 5) but not in the presence of BCL3 D (lanes 4 and 7). Similar results were obtained without BCL3 N-terminal antiserum and with nickel-NTA bead-purified H 6 BCL3 (not shown). This complex pattern of BCL3 activity indicates that phosphorylation may have several effects. The data shown below support a comprehensive model of BCL3 action which is presented under "Discussion." The BCL3 C-terminal Domain Is Not Required for Inhibition-Recent work indicates that the IB C-terminal domain is necessary to inhibit NF-B binding to DNA under some circum- Graded phosphatase treatment and proteolytic digestion reactions were performed as described under "Materials and Methods." A, BCL3 in whole cell extracts was subjected to graded dephosphorylation, and Western blotting was performed with antiserum to the BCL3 ARD. Lanes 1-3 show wild type (wt)BCL3, and lanes 4 -7 show H 6 BCL3. B, BCL3 in whole cell extracts was subjected to graded dephosphorylation and exhaustive digestion with LysC or GluC. C-terminal peptides were analyzed with 14% SDS-PAGE and Western blotting using antibodies to BCL3's C terminus. Undigested BCL3 P (lane 1) and BCL3 D (lane 10) are shown. C, 32 P-labeled H 6 BCL3 was purified on nickel-NTA beads and subjected to partial proteolytic digestion. Peptides were resolved on 14% SDS-PAGE and transfered to a PVDF membrane, and 32 P-labeled peptides were visualized by autoradiography (center panel). The same membrane was then subjected to Western blotting with antibodies to the BCL3 C terminus (left panel) or the BCL3 ARD (right panel). D, 32 P-labeled and nickel-NTA bead-purified H 6 BCL3 (lanes 1-3), which contains a small amount of the 55-kDa phosphoform, and H 6 BCL3⌬N (lanes 4 and 5), which has multiple phosphoforms, were subjected to exhaustive digestion with LysC or GluC. The left panel shows a Western blot, blotted with antibody to the BCL3 C terminus. The right panel is an autoradiograph of the same membrane, performed prior to Western blotting. The autoradiographs are overexposed to detect light bands.
FIG. 2. BCL3 activity is dependent on BCL3 phosphorylation and concentration. Gel mobility shift analyses were performed as described under "Materials and Methods." Each sample has an equal volume of p52-transfected or untransfected nuclear extract and antiserum to the BCL3 N terminus, and each sample is filled to equal volumes of cell extract with untransfected, mock-treated extracts. A, p52 was titrated with control or CIP-treated full-length BCL3. B, p52 was titrated with control or CIP-treated BCL3⌬C. The autoradiograph is overexposed to detect BCL3⌬C-(p52) 2 -B complexes.
stances. The C-terminal domain of mammalian IB␣ (18) and of the avian IB␣ homolog p40 (10) is required for these proteins to inhibit B binding by NF-B. In contrast, the ARD of the Drosophila IB cactus is sufficient for inhibition of the Drosophila NF-B Dorsal (36), and the ARD of bacterially derived BCL3 is sufficient for inhibition of p50 homodimer binding to DNA (4,5). This suggests that different IB proteins may inhibit by distinct mechanisms.
To determine whether or not the C-terminal domain of mammalian cell-derived BCL3 affects its ability to inhibit B binding by (p52) 2 , we performed a gel shift assay in which p52 was titrated with a BCL3 protein truncated at the C-terminal domain (BCL3⌬C) (Fig. 2B). The BCL3⌬C protein contains the entire 7th ankyrin repeat domain and no C-terminal domain residues (7). Full-length BCL3 was incubated with CIP in parallel reactions, and Western blotting was performed to ensure that CIP was active, because CIP treatment does not affect the mobility of BCL3⌬C in SDS-PAGE (not shown).
The results show that BCL3⌬C activity is not affected by CIP treatment and that, at high BCL3⌬C/p52 ratios, BCL3⌬C is an efficient inhibitor. This pattern of concentration-dependent inhibition is similar to that seen with full-length BCL3 D ( Fig. 2A) and suggests that BCL3⌬C and BCL3 D may have similar mechanisms of inhibition. It has been suggested that the C-terminal domain is required for BCL3 to inhibit homodimer B binding, but truncated constructs used in the previous study lacked the 7th ankyrin repeat as well as the C-terminal domain (30). The combined results suggest that phosphorylation at the C-terminal domain of BCL3 impairs its ability to inhibit p52 homodimer binding to B sites.
All Observed Phosphoforms of BCL3 Can Form a BCL3-(p52) 2 -B Complex-There are at least two possible models to explain the mechanism of BCL3 modulation of homodimer binding to DNA that are consistent with results seen in our gel shift assays ( Fig. 2A). One possibility is that only BCL3 P (or certain phosphoisoforms of BCL3 P ) can form a BCL3-(p52) 2 -B complex. Then high concentrations of samples containing predominantly BCL3 D may be required to inhibit homodimer binding to B probe because different BCL3 phosphoforms have different binding affinities for p52 homodimers or other complexes in the binding reaction. In this case, the major species of interest that contain p52 at equilibrium are BCL3 P -(p52) 2 -B, (p52) 2 -B, and BCL3 D -(p52) 2 . A second possibility is that all phosphoforms of BCL3 can form a BCL3-(p52) 2 -B complex, but at high BCL3/p52 ratios a higher order complex is formed that cannot bind to DNA. This higher order BCL3-p52 complex would be predicted to be destabilized by BCL3 phosphorylation, because BCL3 P does not inhibit homodimer binding to DNA. In this case, the major species of interest are BCL3-(p52) 2 -B, (p52) 2 -B, and (BCL3) 2 -(p52) 2 .
To distinguish between these possibilities, it is necessary to show whether 1) BCL3 D is able to form a BCL3 D -(p52) 2 -B complex and 2) BCL3 can form a higher order complex with p52. We studied the formation of a BCL3 D -(p52) 2 -B complex with gel shift analysis and by coprecipitation with tagged B probe and detection by Western blotting. An analysis of the formation of a (BCL3) 2 -(p52) 2 complex is described below (Fig.  4).
First, different phosphoforms of BCL3 were titrated with p52 in a gel shift assay, as shown in Fig. 3A. In the presence of excess p52, BCL3 D forms a BCL3 D -(p52) 2 -B complex which migrates more slowly than the BCL3 P -(p52) 2 -B complex, perhaps due to its decreased net negative charge or a difference in the conformation of the complexes. Under these conditions the BCL3 D -(p52) 2 -B complex is more abundant than the BCL3 P -(p52) 2 -B complex, which suggests that BCL3 P binds relatively weakly to (p52) 2 -B.
To compare the abilities of different phosphoforms of BCL3 to form a BCL3-(p52) 2 -B complex in solution, BCL3 P was subjected to graded dephosphorylation, and mixed phosphoforms of BCL3 were coprecipitated with p52 and a 5Ј-biotinylated B probe immobilized on avidin-coated metal beads and analyzed by Western blotting. Fig. 3B shows Western blots of BCL3 (upper panel) that coprecipitated with immobilized (p52) 2 -B. Precipitated p52 is shown in a Western blot (lower panel). BCL3 P does not coprecipitate efficiently in a BCL3-(p52) 2 -B complex, compared with BCL3 D . (Compare the added BCL3 phosphoforms to those that coprecipitate with (p52) 2 -B.) Thus, BCL3 P associates with (p52) 2 -B relatively weakly, both in solution and in gel shift assays. These results also imply that all observed phosphoforms of BCL3 can form a BCL3-(p52) 2 -B complex.
BCL3 Can Associate with p52 in a Higher Order Complex-Because all observed phosphoforms of BCL3 can form a BCL3-(p52) 2 -B complex, the simplest model for the mechanism of BCL3 inhibition of homodimer binding to DNA is that a higher order BCL3-p52 complex, which cannot bind to DNA, is formed at high BCL3/p52 ratios: BCL3-(p52) 2 -B ϩ BCL3 º (BCL3) 2 -(p52) 2 ϩ B. Thus, BCL3 inhibits homodimer binding to DNA only at high BCL3/p52 ratios. By using a variety of methods, including coimmunoprecipitation, disulfide bond cross-linking, native gel analysis, and glutaraldehyde cross-linking, it has been shown that IB␣ and Cactus associate with NF-B in a 1:2 stoichiometry in vitro and in vivo (18,38,39). Interactions between BCL3 and p50 or p52 homodimers has been less extensively studied, but, as determined by glutaraldehyde crosslinking, bacterially derived BCL3 and p50 can form a 1:2 or a 2:2 complex (4).
To determine whether or not BCL3 forms a higher order complex with p52, we took advantage of the different mobilities of different phosphoforms of BCL3 in SDS-PAGE. We developed a tagged protein sandwich coprecipitation assay in which H 6 BCL3 D , which migrates at an apparent molecular mass of  6). Electrophoresis was performed for 6 h to ensure separation of complexes, because no antiserum was added to supershift BCL3 complexes. B, p52-transfected or untransfected nuclear extract was immobilized with 5Ј-biotinylated B probe on avidin-coated Dynal beads. Immobilized complexes were suspended in binding buffer, aliquoted into sample tubes, and titrated with heterogeneous BCL3 phosphoforms, shown as form A or form B. Complexes were pelleted, and precipitated proteins were eluted by boiling in sample buffer and detected by Western blotting with antibodies to the BCL3 C terminus (above) or to p52 (below). approximately 55 kDa, was immobilized on nickel-NTA beads, while wtBCL3, which is not specifically retained by nickel-NTA beads, was primarily in the phosphorylated, 60-kDa form but had detectable amounts of minor phosphoforms (as shown in Fig. 1A). Immobilized H 6 BCL3 D was titrated with wtBCL3 or CIP-treated wtBCL3 D in the absence or presence of p52; coprecipitated complexes were eluted in a buffer that does not release H 6 BCL3 from beads; and wtBCL3 was detected by Western blotting.
The results of the sandwich coprecipitation assay are shown in Fig. 4. Lanes 4, 8, and 12, with no wtBCL3 present, show that some H 6 BCL3 D elutes from beads, but this background is low. In lanes 5-7, immobilized H 6 BCL3 D was incubated with increasing amounts of wtBCL3 in the absence of p52: here, a low level of background wtBCL3 binding and/or H 6 BCL3 D elution is detectable. In lanes 9 -11, wtBCL3 D was incubated with immobilized H 6 BCL3 D in the presence of p52; here, both elutable wtBCL3 D and immobilized H 6 BCL3 D migrate at approximately 55 kDa and thus cannot be distinguished; however, the bands are much darker than background bands, so it is likely that higher order complexes have formed. In lanes 13-15, untreated wtBCL3 was incubated with immobilized H 6 BCL3 D in the presence of p52: these lanes show that wtBCL3 P coprecipitates with H 6 BCL3 D in the presence of p52. Note that the relative proportion of BCL3 phosphoforms precipitated varied with BCL3 concentration. In particular, the minor 65-kDa form may preferentially form a (BCL3) 2 -(p52) 2 complex.
These results show that BCL3 can associate with p52 in a higher order complex. Previous work has shown that bacterially derived BCL3 can form a 1:2 or a 2:2 complex with p50 proteins by the technique of glutaraldehyde cross-linking, which indicates that direct interactions between the proteins are likely to be responsible for the higher order complex. This indicates that BCL3 inhibits NF-B B binding by a mechanism that is unique among IB proteins (see "Discussion").
BCL3 P Enhancement of p52 Homodimer Binding to B Sites in the Presence of Excess Nontarget DNA-Several groups have observed that IB proteins can enhance B binding by NF-B, and IB␣ can enhance B binding by (p50) 2 and (p52) 2 (9); the avian IB p40, when mutated at acidic residues in its C-terminal domain, can enhance B binding by (c-Rel) 2 (10); and BCL3 P can enhance (p50) 2 binding to a variety of B sites (11). The mechanism of this enhancement has not been elucidated.
One possibility is that BCL3 P enhances B binding by p52 homodimers, at least in part, by increasing the overall rate of their binding to B sites. NF-B proteins have a high nonspecific DNA binding activity, and it has been shown that doublestranded oligonucleotides, such as poly(dI-dC), can reduce the overall rate of NF-B binding to B sites (37). For standard gel shift assays, proteins in cellular extracts are preincubated with nonspecific DNA to reduce probe binding by nonspecific DNAbinding proteins. Homodimers bound to nonspecific DNA must dissociate before they can bind to B probe: (preincubated (p52) 2 The (p52) 2 -B complex forms more slowly in the presence of higher concentrations of nonspecific DNA because there is less free (p52) 2 available to rapidly associate with B probe. We predicted that two conditions are required for BCL3 P to increase the overall rate of homodimer binding to B probe in the presence of excess nonspecific DNA. First, BCL3 P must associate with homodimers and reduce their affinity for DNA, so that there is more free (p52) 2 available to rapidly associate with B probe. Second, BCL3 must dissociate from the BCL3-(p52) 2 -B complex: (preincubated BCL3 P -(p52) 2 High BCL3 P /p52 ratios in binding reactions would be expected to favor the formation of a BCL3 P -(p52) 2 -B complex and make the formation of a (p52) 2 -B complex difficult to study. Thus, to study the effect of BCL3 P on the formation of a (p52) 2 -B complex, we used low BCL3 P /p52 ratios.
Gel shift assays were performed to determine the effect of low concentrations of BCL3 P on the rate of (p52) 2 binding to B sites in the presence of excess nonspecific DNA as shown in Fig.  5A. Proteins were preequilibrated with a 6-fold or a 120-fold molar excess of nonspecific DNA over probe, and the B probe was added and allowed to incubate for decreasing periods before the binding reactions were loaded on the gel. Under these   FIG. 4. BCL3 precipitates in a higher order complex with p52. A tagged protein sandwich coprecipitation assay was performed as described under "Materials and Methods." Briefly, H 6 BCL3 was immobilized on nickel-NTA beads, dephosphorylated with CIP, and then washed. Beads:H 6 BCL3 D were incubated with p52-transfected or untransfected nuclear extract. Beads were pelleted and then incubated with CIP-treated (lanes 9 -11) or untreated (lanes 5-7 and 13-15) wt-BCL3 extract. Precipitated complexes were eluted in a buffer that does not dissociate H 6 BCL3 from beads. Coprecipitated BCL3 was detected by Western blotting with antiserum to the BCL3 ARD. Representative samples of immobilized H 6 BCL3 D are shown in lanes 1 and 2. As a control for the requirement for p52 in a higher order complex, no p52 was added in lanes 4 -7, and as a control for background elution of immobilized H 6 BCL3, no wtBCL3 was added in lanes 4, 8, and 12. Reference lanes (lanes 3 and 16) show equal amounts of CIP-treated wtBCL3 so that the relative extent of exposure can be compared between blots.
FIG. 5. BCL3 P accelerates p52 homodimer binding to B sites in the presence of excess nontarget DNA. 30 min before the addition of 32 P-labeled B probe, proteins were preequilibrated with nonspecific DNA (sonicated salmon sperm DNA) at either 6-or 120-fold molar excess. Standard gel shift assays were performed except that probe was added at different time points before samples were loaded onto the gel. A, each sample contained equal aliquots of BCL3-transfected or untransfected whole cell extracts, p52-transfected nuclear extracts, and BCL3 N-terminal antiserum. B, a gel shift assay with purified control-or CIP-treated BCL3, or purified untransfected extracts, was performed with 30-fold molar excess of nonspecific DNA, as above, except that no BCL3 N-terminal antiserum was added. Samples were incubated with 32 P-labeled B probe for 3 min prior to loading the gel.
conditions, both the off-rate of protein bound to nonspecific DNA and the on-rate of protein binding to B probe are expected to affect the time required for the formation of (p52) 2 -B and BCL3-(p52) 2 -B complexes.
The rate of (p52) 2 binding to the B probe, with 6-fold excess of nonspecific DNA, compared with the rate with 120-fold excess, shows that nonspecific DNA significantly retards B binding by (p52) 2 , consistent with previous results. BCL3 P has little detectable effect in the presence of the lower concentration of nonspecific DNA, and equilibrium is apparently approached after 35 min in both the presence and the absence of BCL3 P . With the higher concentration of nonspecific DNA, however, BCL3 P significantly increases the rate of B binding by (p52) 2 . Quantitative analyses measuring the rate of homodimer binding to B probe in the presence of increasing nonspecific DNA confirmed these results (not shown). Dephosphorylated BCL3, which has not been observed to enhance (p52) 2 binding to B sites in standard gel shift assays, does not accelerate B binding by (p52) 2 in the presence of excess nonspecific DNA (Fig.  5B).
These results indicate that BCL3 P enhances B binding by homodimers, at least in part, by impairing their ability to bind to nontarget DNA. Thus, more of a (p52) 2 -B complex is observed in the presence of BCL3 P in some standard gel shift assays because the binding reactions have approached equilibrium only in the presence of BCL3 P . DISCUSSION In this work, we elucidate the mechanisms of BCL3 modulation of B binding by p52 homodimers. All observed phosphoforms of BCL3 can form a BCL3-(p52) 2 -B complex, but whether BCL3 enhances or inhibits DNA binding by p52 homodimers depends on the concentration and phosphorylation state of BCL3. At high BCL3/p52 ratios, phosphorylated BCL3 has little or no inhibitory activity, whereas dephosphorylated BCL3 is an efficient inhibitor. This suggests that the formation of a higher order BCL3-p52 inhibitory complex depends on BCL3 concentration and phosphorylation. We show that BCL3 coprecipitates with p52 in a higher order complex. Our results support a model in which constitutively phosphorylated BCL3 enhances p52 homodimer binding to B sites, at least in part, first by impairing its binding to DNA and then by forming a weak BCL3-(p52) 2 -B complex from which BCL3 is released.
The physiological role of BCL3 is not understood. BCL3 may activate gene expression either directly by forming a BCL3homodimer-B complex or indirectly by acting as an antirepressor. Our data indicate that the formation of a BCL3-homodimer-DNA complex depends on BCL3 concentration and phosphorylation. BCL3 activity in vivo may potentially be controlled by many factors, including the specific NF-B homodimer, the B sequence, protein phosphorylation, and interactions with other transcription factors. Thus, the physiological role of BCL3 may be highly dependent on local influences.
A Model for the Mechanism of BCL3 Action-Based on our results, we propose a model for the mechanism of BCL3 action which is outlined in Fig. 6. Different aspects of the model are discussed in detail below. Fig. 6A illustrates the key complexes formed in the presence of BCL3. The abundance of each of these complexes at equilibrium is dependent both on BCL3 concentration and phosphorylation. At high ratios of BCL3 to p52, BCL3 phosphorylation destabilizes a higher order complex (reaction 5), whereas BCL3 D inhibits p52 homodimer DNA binding by forming a relatively stable higher order BCL3-p52 complex. At low ratios of BCL3 to p52, BCL3 phosphorylation destabilizes the BCL3-(p52) 2 -B complex (reaction 4). However, BCL3 P does not effectively inhibit p52 homodimer DNA binding in a 1:2 stoichiometry because BCL3 P dissociates from a relatively weak BCL3 P -(p52) 2 -B complex (reaction 3). In summary, BCL3 phosphorylation destabilizes both a BCL3-(p52) 2 -B complex and a (BCL3) 2 -(p52) 2 complex. Fig. 6B illustrates a model for how these combined phosphorylation effects result in BCL3 P -mediated enhancement of homodimer B binding. Because a BCL3 P -(p52) 2 -nonspecific DNA complex is weak, more p52 homodimers are free to associate rapidly with B probe in the presence of BCL3 P . Similarly, a BCL3 P -(p52) 2 -B complex is weak, but because BCL3 P dissociates from the complex, the quantity of (p52) 2 -B is only minimally reduced at equilibrium. BCL3 Phosphorylation-Whether or not BCL3 phosphorylation is regulated in the cell or regulates the ability of BCL3 to act as a transactivator is not known. We and others (11,32) have shown that BCL3 regulation of p50 or p52 homodimer binding to B sites is phosphorylation-dependent. We have shown that all observed phosphoforms of BCL3 are able to form a BCL3-homodimer-B complex (Fig. 3) and that BCL3 phosphorylation has at least two effects: it abrogates the ability of BCL3 to inhibit p52 homodimer B binding ( Fig. 2A), and it impairs BCL3 association in a BCL3-homodimer-B complex ( Figs. 2A, 3, and 5A).
Two lines of evidence suggest that phosphorylation within the C-terminal domain affects BCL3 activity. First, the vast majority of phosphorylation of COS cell-derived BCL3 is within its C-terminal domain (Fig. 1), and a graded change in phosphorylation corresponds to a graded change in BCL3 activity ( Figs. 2A and 3A). Second, the C-terminal domain is not required for concentration-dependent inhibition of p52 homodimer binding to DNA (Fig. 2B). This suggests that Cterminal phosphorylation impairs inhibition by effects on conformation or charge. BCL3 D is sensitive to LysC digestion, FIG. 6. Model of the mechanism of BCL3 action. Numbered reactions shown in A correspond to those shown in B. A, dephosphorylated BCL3 associates in a BCL3-(p52) 2 -B complex (reaction 4) at low concentrations and forms a higher order inhibitory complex (reaction 5) at high concentrations. BCL3 phosphorylation reduces BCL3 association in a higher order complex (reaction 5) and has other effects that result in an enhancement of the rate of homodimer binding to B sites, as illustrated in B. B, when p52 is preincubated with excess nonspecific DNA before the addition of B probe, most (p52) 2 is nonspecifically bound. When B probe is added, the overall rate of (p52) 2 binding is determined by the rates of (p52) 2 dissociation from nonspecific DNA (reaction 1A) and of (p52) 2 association with B (reaction 1B). Relative to (p52) 2 , BCL3 P -(p52) 2 has a lower affinity for DNA, and more BCL3 P -(p52) 2 is free to bind to B probe. Thus, in the presence of BCL3 P , the overall rate of B binding is determined primarily by the rate of association (reaction 4B). Because at low concentrations BCL3 P is largely released from a BCL3 P -(p52) 2 -B complex (reaction 3), the quantity of (p52) 2 -B at equilibrium is minimally reduced. whereas BCL3 P is not, which suggests that the conformation of BCL3 is affected by phosphorylation (not shown). Our data do not rule out the possibility that phosphorylation of N-terminal or ARD residues may also affect BCL3 activity.
BCL3 Inhibition of p52 Homodimer Binding to B Sites-Gel shift analysis shows that BCL3 inhibition of p52 homodimer binding to B sites depends on BCL3 concentration and phosphorylation ( Fig. 2A). Using tagged protein coprecipitation analysis, we have shown that BCL3 can form a higher order complex with p52 (Fig. 4). In combination with previous results that have shown that BCL3 forms a 2:2 complex with p50 (4), the results indicate that the mechanism of BCL3 inhibition of p52 homodimer binding to B sites is through the concentration-and phosphorylation-dependent formation of a 2:2 BCL3-p52 complex. This mechanism of inhibition of NF-B DNA binding is unique among IB proteins, as discussed below.
Mutational analyses have identified three IB contact sites within the Rel homology domain, including the nuclear localization signal. The crystallographic structure of p50 suggests that these sites form a composite surface or "IB-binding cleft" (36, 40 -42). BCL3 may associate with homodimers at this region.
Our results show that, at high ratios of BCL3 to p52, at least one more BCL3 molecule can associate with (p52) 2 . Two rotationally symmetric sites may exist within the IB-binding cleft of p50 and p52 homodimers, and a BCL3 molecule may be able to bind at each site. BCL3 phosphorylation may increase steric hindrance or electrostatic repulsion and decrease the stability of a 2:2 complex, thereby reducing BCL3's ability to inhibit p52 DNA binding.
Our results appear to conflict with previous work that has suggested that phosphorylation increases, rather than decreases, the ability of insect cell-derived BCL3 to inhibit p50 homodimer binding to B sites (32). This contrasting result may be due to differences in the B sequence, the specific NF-B homodimer, BCL3 phosphorylation in insect cells, or binding conditions. We have observed that BCL3 P is an efficient inhibitor in the presence of unphysiologically high concentrations of divalent cation, whereas the activity of BCL3 D is not affected (not shown). Divalent cation may stabilize the formation of a 2:2 BCL3-p52 complex or it may prevent BCL3 P dissociation from a relatively weak BCL3 P -(p52) 2 -DNA complex so that BCL3 P inhibits in a 1:2 stoichiometry.
BCL3 P Enhancement of p52 Homodimer Binding to B Sites-BCL3 has been shown to increase B binding by p50 or p52 homodimers in vitro. Our data show that BCL3 P increases the rate of p52 homodimer binding to B sites in the presence of excess nonspecific DNA (Fig. 5). The simplest explanation is that BCL3 P impairs homodimer binding to both target and nontarget DNA. With excess nontarget DNA, most of the homodimer is bound to nontarget DNA and is not available to bind to B sites rapidly; in the presence of BCL3 P , more homodimer is free to bind rapidly to B sites, and the reaction is driven toward the formation of a homodimer-B complex with the release of BCL3 P (Fig. 6B).
Protease sensitivity and circular dichroism spectroscopy analyses indicate that p50 homodimers change their conformation upon binding to either specific or nonspecific DNA sequences (43,44). It may be that association with BCL3 P impairs conformational adjustments required for (p50) 2 or (p52) 2 to bind DNA.
For BCL3 P to enhance homodimer binding to B sites, it must first associate with homodimers to increase their rate of B binding and then dissociate. Gel shift and tagged B coprecipitation analyses show that, compared with BCL3 D , BCL3 P associates weakly in a BCL3-(p52) 2 -B complex (Fig. 3). High concentrations of BCL3 P favor the formation of a BCL3 P -homodimer-B complex in vitro; but in the nucleus the association of BCL3 P or (p52) 2 with other proteins, or other local influences, may affect the retention of BCL3 P in the complex. BCL3 D does not increase the overall rate of homodimer B binding (Fig. 5B). Our data show that a BCL3-(p52) 2 -B complex is stabilized by BCL3 dephosphorylation. This supports a model in which weak DNA binding and the release of BCL3 from a BCL3-(p52) 2 -B complex are responsible for the ability of BCL3 to enhance homodimer B binding.
RNA polymerase II (45) and the lac repressor (46) have also been shown to have an impaired ability to bind to nontarget DNA as a result of their association with ligands. Whether or not this mode of controlling DNA binding activity is physiologically relevant is not known and is discussed below.
BCL3 Regulation of p50 Homodimers-BCL3 appears to be a more efficient inhibitor with p50 than with p52 homodimers (33), and a BCL3-(p50) 2 -B complex is more difficult to detect than a BCL3-(p52) 2 -B complex (28). Our preliminary results indicate that BCL3 has the same spectrum of activity with p50 homodimers as with p52 and that BCL3 phosphorylation reduces its inhibition of p50 homodimer binding to DNA (not shown). It may be that BCL3 mechanisms of enhancement and inhibition are similar for p50 and p52 homodimers, but a 2:2 complex is more stable or a BCL3-homodimer-B complex is less stable with p50 than with p52 homodimers.
IB Activity in the Nucleus-The possibility that other members of the IB family have a nuclear role in the regulation of NF-B activity is supported by recent studies demonstrating that IB␣ and IB␤, which are primarily cytosolic proteins, can localize in the nucleus and affect NF-B activity (6,51).
There are several lines of evidence that different IB proteins use distinct mechanisms to inhibit NF-B binding to B sites as follows: 1) only BCL3 has been shown to form a higher order (2:2) complex with NF-B subunits, whereas other IBs have been shown to associate with NF-B in a 1:2 stoichiometry; 2) although IB␣ and p40 require acidic residues within their C-terminal domains to inhibit, BCL3 and cactus require only the ARD; and 3) although IB␣ and BCL3 can dissociate preformed NF-B-B complexes, IB␥ cannot (32,52,53).
The ability of IB proteins to form stable IB-NF-B-B complexes is also distinct for different IB proteins; all observed phosphoforms of BCL3 are able to form a BCL3-(p52) 2 -B complex; only dephosphorylated IB␤ forms an IB␤-p50p65-B complex (6), and mammalian IB␣ is apparently unable to form a stable IB-NF-B-B complex under any circumstances.
This diversity in the activity of IB family members suggests that different IB proteins have distinct and synchronous roles in the regulation of nuclear NF-B activity. By enhancing or inhibiting nontransactivating homodimer binding to B sites, BCL3 may coordinate with other IB proteins to affect which NF-B dimers are available to bind to B sites; within a ternary complex, BCL3 may have distinct functions of its own.
The Physiological Role of Enhancement-NF-B dimers bind to a variety of B sequences with exceptionally high affinity (9,37), and transactivation activity by different dimers is highly sensitive to minor changes in the B sequence (47,48). Thus, the specificity of NF-B-controlled gene expression is determined largely by which NF-B dimers are available to bind to B sites. If (p50) 2 and (p52) 2 are constitutively available to compete for the occupation of B sites, they may impair the ability of transactivating NF-B proteins to affect gene expression. Thus, it was originally proposed that a primary role for BCL3 may be to inhibit homodimer binding to B sites and act as an antirepressor. In light of this, it is surprising that thy-mocytes that express a BCL3 transgene show an increase, rather than a decrease, in B binding by p50 homodimers in vitro, with no change in the level of the p50 subunit in the nuclear extract (11). This is consistent with the notion that BCL3 enhancement of homodimer binding to B sites may be a physiologically relevant activity.
Recently, it has been shown that BCL3 can increase the quantity of nuclear p50 homodimers by inducing a redistribution of the subunits of cytoplasmic heterodimers containing p50 and its precursor p105 to yield p50 homodimers that translocate to the nucleus; this occurs without an effect on processing of p105 (54). In combination with our results, this suggests that BCL3 may inhibit NF-B-dependent gene expression by increasing nuclear levels of nontransactivating NF-B homodimers as well as by increasing their ability to bind to B sites.
There are several other mechanisms by which homodimer binding to available B sites may be regulated in vivo. p50 and p52 subunits are apparently not phosphorylated in the cell (49), and there is no evidence that their activity is affected by phosphorylation. Homodimer B binding may be affected by a monomer/dimer equilibrium, association with specific inhibitors, cooperative binding interactions with other transcription factors, or association with chromatin. The extent to which DNA is available to act as a nonspecific competitor in vivo is not known, but recently it has been demonstrated that several sequence-specific DNA-binding proteins have a general affinity for interphase chromatin (50). Thus, p50 or p52 homodimer binding to bulk chromatin may impair their binding to B sites. Further studies must be performed to determine whether or not NF-B associates nonspecifically with chromatin in the cell.
BCL3 P which remains associated in a BCL3 P -homodimer-B complex may act as a transactivator, but BCL3 P which is released from the complex may function solely to escort (p50) 2 and (p52) 2 to B sites. IB␣, which has also been observed to enhance p50 or p52 homodimer binding to B sites in vitro, may have an escort function as well. An IB escort function would be predicted to increase the ability of p50 or p52 homodimers to compete for B site occupancy and thus decrease the efficiency or the duration of B binding by transactivating NF-B dimers.