Mechanism of IκBα Binding to NF-κB Dimers*

X-ray crystal structures of the NF-κB·IκBα complex revealed an extensive and complex protein-protein interface involving independent structural elements present in both IκBα and NF-κB. In this study, we employ a gel electrophoretic mobility shift assay to assess and quantitate the relative contributions of the observed interactions toward overall complex binding affinity. IκBα preferentially binds to the p50/p65 heterodimer and p65 homodimer, with binding to p50 homodimer being significantly weaker. Our results indicate that the nuclear localization sequence and the region C-terminal to it of the NF-κB p65 subunit is a major contributor to NF-κB·IκBα complex formation. Additionally, there are no contacts between the corresponding nuclear localization signal tetrapeptide of p50 and IκBα. A second set of interactions involving the acidic C-terminal/PEST-like region of IκBα and the NF-κB p65 subunit N-terminal domain also contributes binding energy toward formation of the complex. This interaction is highly dynamic and nonspecific in nature, as shown by oxidative cysteine cross-linking. Phosphorylation of the C-terminal/PEST-like region by casein kinase II further enhances binding.

X-ray crystal structures of the NF-B⅐IB␣ complex revealed an extensive and complex protein-protein interface involving independent structural elements present in both IB␣ and NF-B. In this study, we employ a gel electrophoretic mobility shift assay to assess and quantitate the relative contributions of the observed interactions toward overall complex binding affinity. IB␣ preferentially binds to the p50/p65 heterodimer and p65 homodimer, with binding to p50 homodimer being significantly weaker. Our results indicate that the nuclear localization sequence and the region C-terminal to it of the NF-B p65 subunit is a major contributor to NF-B⅐IB␣ complex formation. Additionally, there are no contacts between the corresponding nuclear localization signal tetrapeptide of p50 and IB␣. A second set of interactions involving the acidic C-terminal/PEST-like region of IB␣ and the NF-B p65 subunit N-terminal domain also contributes binding energy toward formation of the complex. This interaction is highly dynamic and nonspecific in nature, as shown by oxidative cysteine cross-linking. Phosphorylation of the C-terminal/ PEST-like region by casein kinase II further enhances binding.
The Rel/NF-B family transcription factors perform a vital role in mediating the cellular response to stress, inflammation, the immune response, and apoptosis (1)(2)(3)(4). Mammalian Rel/ NF-B family polypeptides include p65 (RelA), p50, p52, RelB, and the proto-oncoprotein c-Rel. These subunits associate in various combinations to form homodimers and heterodimers with distinct but overlapping functions. Among the most abundant and best understood of these dimers are the p50/p65 heterodimer and the homodimers of p50 and p65 (5). The functions of other NF-B dimers are generally restricted to specific cell types.
In most cells the NF-B p50/p65 heterodimer exists in the cytoplasm as a complex with the inhibitor protein IB␣. IB␣ inactivates the NF-B p50/p65 heterodimer by masking the NF-B nuclear localization sequences (6). Activation of signaling pathways by extracellular signals, such as tumor necrosis factor or interleukin-1, leads to proteolysis of IB␣, allowing active NF-B p50/p65 heterodimer to translocate into the nucleus (7).
IB␣ is composed of three distinct regions: an N-terminal signal receiving domain (SRD), 1 a central ankyrin repeat-containing domain (ARD), and an acidic C-terminal/PEST-like region rich in the amino acids proline, glutamic acid, serine, and threonine (8). IB␣ is phosphorylated my multiple proteins in vivo. The C-terminal/PEST-like region in constitutively phosphorylated by casein kinase II (CKII) (9,10), and the SRD contains sites of inducible, signaling-dependent phosphorylation (11). X-ray crystal structures of IB␣ in complex with the NF-B p50/p65 heterodimer reveal an extensive and complex protein-protein interface (12,13). The structures reveal that an apparently flexible segment of the NF-B p65 subunit, which contains the NLS, becomes ordered upon complex formation and makes several contacts with IB␣. Interestingly, no direct contact is observed between the corresponding segment of the NF-B p50 subunit and IB␣. The structures also reveal that the N-terminal domain of the p65 subunit contacts the Cterminal/PEST-like region of IB␣ and its own dimerization domain through long range electrostatic interactions. Included in this interaction are many of the amino acid residues that mediate NF-B-DNA contacts. These electrostatic interactions and the accompanying conformational change observed in the NF-B p65 subunit were proposed to provide the DNA-inhibitory binding function of IB␣.
Previous experiments have been carried out to define regions of NF-B dimers that are important for IB␣ binding (14). For these previous experiments, we made use of a fluorescence polarization competition assay to assess the roles of different deletion fragments of NF-B and IB␣. However, this approach limited our studies to NF-B deletion protein constructs that display appreciable DNA binding affinity. As a result, we were not able to assess the contribution to overall complex binding affinity by regions of the NF-B dimers independent of their DNA binding sequences.
In this report we make use of a gel electrophoretic mobility shift assay to isolate and characterize the interactions observed in the NF-B⅐IB␣ x-ray crystal structures. Our results show that: 1) Escherichia coli expressed full-length IB␣ binds the NF-B p50/p65 heterodimer with the highest affinity of the three NF-B dimers tested. IB␣ binding affinity toward the p50 and p65 homodimers is 20-and 2.5-fold lower, respectively, than toward the p50/p65 heterodimer. 2) The segment C-terminal to the NF-B p65 subunit NLS contributes 5-6-fold to the overall binding affinity both in the case of the p65 homodimer and the p50/p65 heterodimer. By contrast, the corresponding segment of p50 contributes only 2-fold to IB␣ binding affinity. 3) Within the context of the NF-B p50/p65 heterodimer, neither the N-terminal domain nor the NLS of p50 contributes to IB␣ binding. In contrast, the corresponding segments of the p65 subunit do make significant contributions to overall binding affinity. 4) When compared with its unphosphorylated form, casein kinase II-phosphorylated IB␣ binds the p50/p65 heterodimer and the p65 homodimer with 33-and 9-fold higher binding affinity, respectively. 5) Finally, we report oxidative cross-linking data that illustrate the dynamic nature of the electrostatic interactions between the acidic Cterminal/PEST-like region of IB␣ and NF-B.

MATERIALS AND METHODS
Cloning and Purification of RS-tagged IB␣-The RS-IB␣ clone was created by ligating together three components. The components included one IB␣ construct corresponding to amino acid residues 1-317, 1-302, or 1-287, the RS tag, and the pET15b vector (Novagen). The N-terminal primer used for IB␣ amplification by polymerase chain reaction was: 5Ј-TAGCTAGTCGACTTCCAGGCGGCCGAGCGC-3Ј. The SalI restriction site is underlined. The C-terminal primers all contained a BamHI restriction site (underlined) and had the following sequences: 317, 5Ј-TGCAGAGGATCCTCATAACGTCAGACGCTGGC-C-3Ј; 302, 5Ј-GAAGTCGGATCCCTACTCGTCCTCTGTGAACTC-3Ј; and 287, 5Ј-GAAGCTGGATCCCTACTCCTCATCCTCACTCTC-3Ј. The RS tag was generated by annealing the two oligonucleotides with the following sequences: 5Ј-TATGAGAGATGCTCCTCGCGAAAGATCAC-CAACCACGC-3Ј and 3Ј-ACTCTCTACGAGGAGCGCTTTCTAGTGGT-TGGTGCGAGCT-5Ј. The resulting double-stranded DNA was phosphorylated on its 5Ј-hydroxyl groups with polynucleotide kinase (New England BioLabs) prior to ligation with the IB␣ insert and vector. The amino acid sequence encoded for by the RS tag oligonucleotide is MRDAPRERSPTRLE. This represents the specific phosphorylation site for Sky1 (the splicing kinase in yeast 1) as derived from its substrate Npl3 (the phosphorylated serine is underlined) (15). The pET15b vector additionally encodes an N-terminal hexahistidine tag, which was used to purify the protein using a Ni 2ϩ affinity resin (His-Bind, Novagen). The protein was further purified by size exclusion chromatography on a Superdex 75 column (Amersham Pharmacia Biotech).
Phosphorylation of RS-IB␣-Phosphorylation of each of the RS-IB␣ constructs (1-317, 1-302, and 1-287) was carried out using either the enzyme Sky1 (a gracious gift from Brad Nolen) alone or in combination with casein kinase II (CKII; New England Biolabs). In both cases the buffer used was 20 mM Tris-HCl (pH 7.5), 50 mM KCl, and 10 mM MgCl 2 . RS-IB␣ was added to 50 M in 100-l reactions with 0.2 mCi of [ 32 P]ATP. For the Sky1-only reactions, the kinase was added to 2 nM and incubated at 30°C for 30 min. In reactions containing both Sky1 and CKII, 200 units of CKII was added to the Sky1 reactions along with 200 M cold ATP and incubated at 30°C for an additional 30 min. Excess [ 32 P]ATP was then removed, and the buffer was exchanged by dilution in 20 mM Tris-HCl (pH 7.5), 50 mM NaCl, 2 mM EDTA, and 2 mM dithiothreitol followed by concentration in a centricon-10 (Amicon). Concentrated protein (100 M) was then flash frozen in liquid nitrogen and stored at Ϫ80°C.
Electrophoretic Mobility Shift Assays-EMSAs with RS-IB␣ 1-317 were run with each of the NF-B dimers listed in Table I. The p50 39 -376 / p65 19 -325 heterodimer was employed in EMSAs for truncated RS-IB␣ proteins and CKII-labeled RS-IB␣. Following labeling, RS-IB␣ at 1 nM concentration was titrated with increasing NF-B in 20-l reactions containing 20 mM Tris (pH 7.5), 100 mM NaCl, 1 mM dithiothreitol, 0.01% Triton X-100, 50 g/ml bovine serum albumin, and 5% glycerol (v/v). There was one exception; reactions between the CKII-labeled RS-IB␣ and the p50 39 -376 /p65 19 -325 heterodimer were performed at 0.1 nM IB␣. The reactions were allowed to equilibrate at room temperature for 30 min and then loaded on 0.25ϫ TBE gels. 6% polyacrylamide gels were used for all reactions except those containing NF-B dimers lacking both N-terminal domains in which 7.5% polyacrylamide gels were used. After running for 5 h at 140 V constant, the gels were dried and exposed to a phosphor storage plate for 24 h. The plates were then scanned using a Molecular Dynamics Storm 860 scanner and quantified using ImageQuant version 1.2 (Molecular Dynamics).
Mutagenesis of IB␣ and p65 for Oxidative Cysteine Cross-linking-Cysteine mutants of IB␣ and one of the p65 mutants were generated with a single step of polymerase chain reaction. The other cysteine mutants of p65 were generated using a two-step polymerase chain reaction strategy. DNA oligonucleotide primers (listed below) were used to amplify the mutated gene construct, which was subsequently subcloned into the NdeI and BamHI sites of pET11a (p65 mutants) or pET15b (IB␣ mutant) vector. Point mutations were verified by DNA sequencing (DNA sequencing was performed by the Molecular Pathology Shared Resource, UCSD Cancer Center, which is funded in part by NCI Cancer Center Support Grant 5P0CA23100-16). p65 subunits bearing cysteine mutations were purified according to established protocols (16), and the IB␣ 67-288 S288C mutant protein was purified in the same manner as RS-IB␣.
Oxidative Cysteine Cross-linking of IB␣ and NF-B Proteins-Proteins were reduced in 25 mM dithiothreitol and incubated on ice for 1 h. Dithiothreitol was removed by a protein buffer exchange G-25 spin column (Amersham Pharmacia Biotech). Thereafter, 0.5 mg/ml of NF-B and 0.25 mg/ml of IB␣ proteins (approximately equal molar amounts) were incubated in PBS at room temperature for 30 min. Nonreducing gel loading buffer was added prior to running on 10% SDS-polyacrylamide gels. Protein was visualized by Coomassie staining, and band shifts were verified by Western blots with antibodies toward both p65 and IB␣.

RESULTS
Experimental Design-We have shown previously via competition assay that within the context of the NF-B p50/p65 heterodimer, the dimerization domain only of p50 as well as the of N-terminal and dimerization domains and the NLS of p65 participate in complex formation with IB␣. We have also shown that the SRD of IB␣ does not contribute to overall NF-B binding affinity (14). The x-ray crystal structures of the IB␣ in complex with the NF-B p50/p65 heterodimer support these biochemical findings. Furthermore, the crystal structures revealed a role for a sequence of 17 amino acids immediately C-terminal to the NF-B p65 subunit NLS (12) as well as identifying the probable mode of interaction between the Cterminal/PEST-like region of IB␣ and NF-B (13).
To quantitate the relative contributions of each of these segments toward overall NF-B⅐IB␣ binding affinity, we employed a gel EMSA. EMSAs were performed by titrating increasing concentrations of NF-B dimer protein constructs against a constant concentration of IB␣. For this assay, an IB␣ fusion protein was prepared containing a peptide derived from the yeast protein Npl3, a natural substrate for Sky1 (15). The tag was placed at the IB␣ N terminus, allowing for 32 P labeling to occur at a location that does not interact with NF-B. IB␣ is a substrate for multiple kinases including CKII (Fig. 1). In the absence of the SR tag, however, Sky1 fails to transfer phosphate to IB␣ (data not shown). The many NF-B homodimer and heterodimer constructs employed in this study (Table I) were engineered to address the contribution of each of the structural elements toward IB␣ complex formation. Fig. 2 shows a schematic representation of the p50 and p65 NF-B monomers and IB␣ and defines the specific elements we targeted.
Residues C-terminal to the p65 NLS Are Essential for IB␣ Binding-Crystal structures of the NF-B⅐IB␣ complexes have defined the segments of the p50 and p65 subunits that are involved in IB␣ binding. The structure reveals that IB␣ binds to the NF-B p65 subunit within the region encompassing residues 19 -319. The homologous segment in p50 includes residues 39 -376 and corresponds to the entire p50 subunit through the NLS polypeptide. The crystal structures further reveal that, within the context of the NF-B p50/p65 heterodimer, only the dimerization domain of p50 (residues 245-350) contacts IB␣. We have constructed and purified a slightly longer version of p65 ending at residue 325 for use in our binding experiments. We also constructed the homologous p50 homodimer and the p50/p65 heterodimer.
EMSAs were carried out on each of these NF-B dimer constructs ( Fig. 3 and Tables II-IV). As expected, IB␣ binds to the p50/p65 heterodimer with highest affinity. The observed equilibrium binding constant is approximately 6 nM, whereas the p65 homodimer exhibits a roughly 2-3-fold lower affinity. We observe that a diffused intermediate band runs between the free IB␣ and IB␣/p65 homodimer complex bands (Fig. 3B). This intermediate band likely represents a complex of the p65 monomer and IB␣. Such an intermediate is observed only in binding experiments including the p65 homodimer and is, perhaps, related to the fact that the p65 subunit forms a significantly weaker dimer than do either the p50 homodimer or the FIG. 1. Sky1 and CKII phosphorylate RS-IB␣ in vitro. SDSpolyacrylamide gel electrophoresis analysis and autoradiography confirms that the kinases used in this study are capable of phosphorylating each of the three RS-IB␣ protein constructs (ending at residues 287, 302, and 317).

TABLE I NF-B constructs used in EMSAs with RS-IB␣
The following constructs were all used in EMSAs to determine their affinities for full length IB␣  . Additionally the full-length NF-B heterodimers were used to determine the effects of C-terminal IB␣ truncation. See Fig. 2 2. Schematic representations of p50, p65, and IB␣. The Rel homology regions of p50 and p65 are represented schematically. The numbering corresponds to the boundaries of the various NF-B constructs used. For p65, two different N termini (1 and 19) and C termini (321 and 325) are used interchangeably because they bear no effects on IB␣ binding. IB␣ contains three distinct regions the SRD, ARD, and C-terminal/PEST-like region. Arrows indicate the borders of the constructs tested. Also indicated is the portion of the PEST-like region containing the five CKII phosphorylation sites.  Fig. 2. Note that the binding of p50 is much weaker than p65 or p50/p65. Also visible is the intermediate band on the p65 gel that is highlighted by asterisks. p50/p65 heterodimer. 2 The p50 homodimer binds IB␣ weakly with an equilibrium dissociation constant of 218 nM.
The structure of the NF-B⅐IB␣ complex solved by Jacobs and Harrison (12) revealed that the C-terminal 17 residues of p65 mediate several contacts with IB␣. To assess and quantitate the importance of these interactions in complex formation, we constructed a p65 homodimer ending at residue 304. The analogous construct of p50, corresponding to amino acids 39 -363, was prepared as was the p50 39 -363 /p65 191-304 NF-B heterodimer. EMSA results indicate that both the p65 homodimer and the p50/p65 heterodimer exhibit a 5-6-fold loss of binding affinity for IB␣ upon removal of the segments Cterminal to the p65 NLS. By contrast, removal of this segment from the p50 homodimer results in a 2-fold loss in IB␣ binding (Tables II-IV).
Role of the NF-B N-terminal Domains in IB␣ Binding-We tested for the role of the NF-B N-terminal domains in binding IB␣. For this purpose, we constructed NF-B p50/p65 heterodimers with one of the NF-B N-terminal domains removed. The p50 245-363 /p65 19 -304 heterodimer binds IB␣ with an affinity similar to that of p50 39 -363 /p65 19 -304 heterodimer, suggesting that removal of the p50 N-terminal domain results in no change in overall binding (Table II). On the contrary, the p50 39 -363 /p65 191-304 heterodimer exhibited an approximately 3-fold decreased affinity toward IB␣, implicating a role for the p65 subunit N-terminal domain in the complex interaction. These data correspond with the documented preference of IB␣ toward the NF-B p65 subunit (Fig. 4).
We measured affinities of the N-terminal truncated proteins in the context of their residues C-terminal to the NLS. Addition of p50 and p65 subunit C-terminal extensions produced little significant changes (Table II). Only a slight decrease in affinity was observed in the case of p50 39 -376 /p65 191-325 with the p65 N-terminal domain was removed. Surprisingly, deletion of the p50 N-terminal domain resulted in 3-4-fold tighter binding for p50 245-375 /p65 19 -325 as compared with the full p50 39 -376 / p65 19 -325 heterodimer. Similar results were obtained for the p65 191-325 homodimer with both the N-terminal domains removed. Specifically, the p65 191-325 homodimer binds IB␣ 2.5fold tighter than does the homodimer p65 19 -325 (Table III). Removal of both of the N-terminal domains of the p50 homodimer causes a slight (2-fold) decrease in its affinity for IB␣ (Table IV). We also tried to measure the IB␣ binding affinity of p50 and p65 homodimers with both the N-terminal domains and the C-terminal extensions removed (p50 245-363 and p65 191-304 ). No quantifiable shift was observed, indicating that IB␣ binds very poorly to these p50 and p65 homodimers (data not shown).
IB␣ Binds One of the Two NLSs of NF-B p50/p65 Heterodimer-Both x-ray NF-B⅐IB␣ complex crystal structures indicated that the NF-B p50 subunit NLS does not contact IB␣ (Fig. 5). To test whether this is indeed the case in solution, we created three different NF-B heterodimers containing progressively shorter p50 subunit NLS polypeptide regions. The p65 NLS polypeptide remained unchanged in each case. These constructs were prepared in the context of the dimerization domains only of p50 and p65 subunits (starting at residues 245 and 191, respectively). Our results show that the presence of the NF-B p50 subunit NLS in the p50 245-363 /p65 191-325 does not enhance affinity of the heterodimer for IB␣ when compared with the p50 245-350 /p65 191-325 heterodimer that lacks the NLS (Table V). We do, however, observe a 2-fold increase in binding affinity for the p50 245-376 /p65 191-325 heterodimer as compared with each of the other two, suggesting that amino acid residues C-terminal to the NLS of p50 may contact IB␣   The p50/p65 heterodimer binds more tightly to IB␣ than both the corresponding p50 and p65 homodimers. K obs is the average observed dissociation from a minimum of three independent experiments. The reported errors represent one standard deviation from the average value.   19 -304 (A) and p50 39 -363 / p65 191-304 (B) heterodimers is shown. The removal of the p50 N-terminal domain has no effect on complex, whereas the p65 N-terminal domain is involved in contacting IB␣. without involving the p50 NLS itself.
Casein Kinase II Phosphorylation at the C-terminal/PESTlike Region of IB␣-The acidic C-terminal/PEST-like region of IB␣ is constitutively phosphorylated in vivo by CKII (9,10). This phosphorylation occurs on some or all of the three serines and two threonines that match the kinase consensus sequence. The x-ray crystal structures of the NF-B⅐IB␣ complex revealed that this region of IB␣ is involved in contacting an extensive basically charged region of the p65 subunit N-terminal domain. We endeavored to understand what effects CKII phosphorylation of IB␣ would have on complex formation. Using the same full-length IB␣ construct as in the experiments previously described, EMSAs were performed with IB␣ phosphorylated by both Sky1 and CKII. Our experiments show that IB␣ phosphorylation by CKII can have a profound effect on NF-B⅐IB␣ complex formation. We tested both the p65 1-325 homodimer and the p50 39 -376 /p65 19 -325 heterodimer (Fig. 6). In both cases the affinity of the complexes was greatly enhanced, 9-fold stronger for p65 homodimer and 33-fold for the heterodimer (Table VI).
Further testing of the of the C-terminal/PEST-like region of IB␣ in both its unphosphorylated and phosphorylated states revealed roles for this region in NF-B binding. We observe that unphosphorylated IB␣ truncated at its C terminus to either residue 287 or 302 exhibits similar NF-B binding affinity as the full-length protein. We have shown previously that IB␣ 67-277 fails to dissociate DNA-bound NF-B as judged by our competition assay. Taken together, these data suggest that a minimal length IB␣ required for binding and dissociation of DNA-bound NF-B is IB␣ 67-287 . Phosphorylation of IB␣ 1-287 by CKII does not enhance affinity toward NF-B p50/p65 heterodimers, and only marginal enhancement is observed in IB␣ 1-302 upon CKII phosphorylation (Table VI). On the other hand, the affinity of full-length IB␣ toward NF-B increases dramatically after phosphorylation with CKII.
The C-terminal/PEST-like Region of IB␣ Participates in Dynamic Interactions with NF-B-The NF-B⅐IB␣ complex x-ray crystal structures reveal that the NF-B p65 subunit adopts a conformation profoundly altered from that exhibited in its DNA-bound complex. The exact position of the N-terminal domain differs slightly in the two NF-B⅐IB␣ complex structures. Furthermore, the average temperature (B) factors of these p65 N-terminal domains are high. Both these observations suggest that the N-terminal domain of the NF-B p65 subunit may participate with the IB␣ PEST-like region in a dynamic interaction driven primarily through long range electrostatic contacts. Although intrigued that this significant change in NF-B conformation could account for the DNA inhibitory binding activity observed upon binding to IB␣, the authors were careful to note that such an altered protein structure may, in fact, represent a crystal-packing artifact (12,13).
In an effort to test whether the acidic C-terminal/PEST-like region of IB␣ is, in fact, capable of such dynamic interactions with NF-B, we performed the following oxidative cross-linking experiment. A mutant IB␣ protein was prepared with serine-288, a side chain located within the C-terminal/PEST-like region, converted to a cysteine. This IB␣ 67-288 S288C protein containing the donor cysteine was incubated with three p65 homodimers, p65 19 6. CKII phosphorylation of IB␣ is required for maximum affinity for NF-B. Gel shift assays for the p50 39 -376 /p65 19 -325 heterodimer with IB␣ with (A) and without (B) CKII phosphorylation. Arrows indicate the positions of the bound and free IB␣; the diagrams next to the gels correspond to those in Fig. 2. The bands in the with CKII phosphorylation gel are more dispersed because of multiple phosphorylation states of IB␣. It was not possible to quantitatively phosphorylate all of the CKII sites in the C-terminal/PEST-like region resulting in multiple species of IB␣. reveal that Arg 246 and Lys 221 are located 11 Å apart from each other on the bottom of the p65 dimerization domain. Glu 22 is positioned more than 20 Å from both Arg 246 and Lys 221 in the N-terminal domain of p65. We observe that IB␣ 67-288 S288C efficiently cross-links with p65 19 -325 K221C and p65 19 -325 -R246C. IB␣ 67-288 S288C also cross-links, albeit with significantly lower efficiency, with the p65 19 -325 E22C homodimer (Fig. 7).

DISCUSSION
The protein-protein interactions at the interface between the NF-B p50/p65 heterodimer and IB␣ are complicated in nature. This complexity arises from the involvement of multiple modular and flexible components at the complex interface: namely, the ARD and C-terminal/PEST-like region of IB␣, and the N-terminal domains, dimerization domains, and the NLS polypeptides of the NF-B dimers. We endeavored to quantitate the contributions made by each of these segments toward complex formation. In addition, we attempted to elucidate how CKII phosphorylation of the C-terminal/PEST-like region of IB␣ modulates its binding to NF-B dimers. Finally, we investigated the nature of the interaction between the NF-B p65 subunit N-terminal domain and the acidic C-terminal/PEST-like region of IB␣.
Our results demonstrate that the interactions between IB␣ and NF-B dimers can be divided into two classes. As the x-ray crystal structures of the NF-B⅐IB␣ complex would clearly indicate, the NLS polypeptide of p65 contributes significantly toward IB␣ binding affinity. The primary sequence of the NF-B p50 and p65 subunit differ significantly in this region which might explain, at least in part, the selectivity of IB␣ toward p65. We have also shown that in the p50/p65 heterodimer, the p50 NLS polypeptide, but not the NLS itself (amino acid residues 360 -363), contributes to IB␣ binding affinity. This observation is particularly interesting in light of several recent reports indicating that the inactive NF-B p50/ p65 heterodimer in complex with IB␣ shuttles continuously between the cytoplasm and nucleus (20,21). It would seem that although nuclear exit of this complex is mediated by a leucinerich nuclear exit sequence within the IB␣ SRD (amino acids 44 -55) (22), constitutive nuclear import could depend on the partially unmasked p50 NLS.
Our experiments further demonstrate a second class of NF-B⅐IB␣ interactions involving the acidic/PEST-like region of IB␣ and the NF-B p65 subunit N-terminal domain. We show that although these contacts contribute to a lesser extent toward overall NF-B⅐IB␣ complex binding affinity, they probably affect significant functional consequences. We characterized these interactions as dynamic in nature. Our cross-linking experiments show that the cysteine mutant IB␣ C-terminal/ PEST-like region forms covalent bonds with a distance-depend-ent efficiency at three different positions on the p65 subunit covering a total distance of greater than 20 Å.
We observe that the N-terminally truncated p50/p65 heterodimer or p65 homodimer bind IB␣ with higher affinities than the respective full-length dimers. This suggests that the absence of the N-terminal domains allows the acidic/PEST-like region of IB␣ more freedom of movement, enhancing its ability to mediate further nonspecific interactions. Finally, we also tested which residues at the C terminus of IB␣ are essential for NF-B binding. We observe that in the absence of C-terminal phosphorylation by CKII the boundary is roughly residue Ser 287 . However, this boundary changes when the acidic PEST is phosphorylated by CKII. It appears that the entire C terminus is required for full binding activity once the PEST-like region of IB␣ is phosphorylated.
The NF-B⅐IB␣ x-ray crystal structures indicate that this second interaction that we have described juxtaposes two large, oppositely charged faces. One contains the acidic IB␣ PESTlike region, and the other involves the basic DNA binding surfaces of the p65 N-terminal domain in a conformation radically different from that observed in its DNA-bound crystal structures. Though we still cannot rule out crystal packing interactions as a contributor to this observed "closed" conformation of the IB␣-bound NF-B p65 subunit, these data support our proposed model of NF-B DNA binding inhibition by electrostatic interference and long range, dynamic interactions with IB␣.  7. The C-terminal/PEST-like region of IB␣ interacts dynamically with the dimerization and N-terminal domains of NF-B. Nonreducing SDS-polyacrylamide gel electrophoresis analysis of oxidative cysteine cross-linking studies of the interaction of IB␣ with p65 homodimer. Lanes are labeled by the constituents of the individual cross-linking reactions. Controls using p65 1-325 and IB␣ 67-302 were run alongside the reactions of the cysteine mutants of p65 and IB␣ 67-288 S288C. The IB␣ 67-288 S288C protein runs at a higher molecular mass that the IB␣ 67-302 because it contains a hexahistidine tag. The arrow indicates the location of the cross-linked p65/ IB␣ complex.