p105·IκBγ and Prototypical IκBs Use a Similar Mechanism to Bind but a Different Mechanism to Regulate the Subcellular Localization of NF-κB

p105, also known as NF-κB1, is an atypical IκB molecule with a multi-domain organization distinct from other prototypical IκBs, like IκBα and IκBβ. To understand the mechanism by which p105 binds and inhibits NF-κB, we have used both p105 and its C-terminal inhibitory segment known as IκBγ for our study. We show here that one IκBγ molecule binds to NF-κB dimers wherein at least one NF-κB subunit is p50. We suggest that the obligatory p50 subunit in IκBγ·NF-κB complexes is equivalent to the N-terminal p50 segment in all p105·NF-κB complexes. The nuclear localization signal (NLS) of the obligatory p50 subunit is masked by IκBγ, whereas the NLS of the nonobligatory NF-κB subunit is exposed. Thus, the global binding mode of all IκB·NF-κB complexes seems to be similar where one obligatory (or specific) NF-κB subunit makes intimate contact with IκB and the nonobligatory (or nonspecific) subunit is bound primarily through its ability to dimerize. In the case of IκBα and IκBβ, the specific NF-κB subunit in the complex is p65. In contrast to IκBα·NF-κB complexes, where the exposed NLS of the nonspecific subunit imports the complex to the nucleus, p105·NF-κB and IκBγ·NF-κB complexes are cytoplasmic. We show that the death domain of p105 (also of IκBγ) is essential for the cytoplasmic sequestration of NF-κB by p105 and IκBγ. However, the death domain does not mask the exposed NLS of the complex. We also demonstrate that the death domain alone is not sufficient for cytoplasmic retention and instead functions only in conjunction with other parts in the three-dimensional scaffold formed by the association of the ankyrin repeat domain (ARD) and NF-κB dimer. We speculate that additional cytoplasmic protein(s) may sequester the entire p105·NF-κB complex by binding through the death domain and other segments, including the exposed NLS.

The NF-B family of transcription factors plays an important role in a large number of cellular processes including immune response and inflammation, cellular development, and differentiation (1)(2)(3)(4). This family comprises of five distinct members, p50, p65, p52, c-Rel, and RelB, that exhibit a high degree of sequence homology at their N termini. This region, known as the Rel homology region, is responsible for important functions like DNA binding, dimer formation, nuclear localization, and IB binding (1)(2)(3)(4).
NF-B dimers are regulated by inhibitor IB proteins, which include IB␣, IB␤, IB⑀, IB␥, p105 (NF-B1), p100 (NF-B2), Bcl-3, IB, and MAIL (4 -21). In most cells NF-B remains inactive as a complex with IB. In response to a variety of extracellular signals, the IB molecule is phosphorylated by IB kinases, which leads to the ubiquitination and subsequent degradation of IB by the proteosome machinery within the cell (22).
It has long been thought that all IB proteins inhibit NF-Bs by masking the nuclear localization signal (NLS) 1 of NF-B and thereby sequestering them in the cytoplasm (1)(2)(3)(4)22). However, recent studies have indicated that the IB proteins differ with regard to their regulation of NF-B subcellular localization. The IB␣⅐NF-B complex exhibits dynamic shuttling between the cytoplasm and the nucleus (23)(24)(25)(26)(27). However, despite its transient presence in the nucleus, NF-B remains bound to IB␣, and this prevents DNA binding and activation of transcription. IB␤, on the other hand, sequesters NF-B in the cytoplasm of resting cells (26 -28). The detailed mechanism of this process is not clear. It has been suggested that other ancillary proteins may be required for this function. 2 p105 and p100 are the precursors of NF-B subunits p50 and p52, respectively, which are located in their N termini. Both p105 and p100 have similar structural organizations (3) (Fig.  1). The central portion of these molecules has a glycine-rich region that has been shown to play a critical role in processing of the precursor (29,30). The C termini resemble other IB molecules that possess ankyrin repeats (AR) (14,15,31). These two proteins also contain a death domain immediately C-terminal to the ankyrin repeat domain (ARD). The IB kinase phosphorylation sites are located further downstream within a region called the destruction box (see Fig. 1). A separate gene also encodes the C-terminal part of p105. This gene product, known as IB␥, has been shown to exist only in certain cell types, like mouse pre-B cells (12) (Fig. 1).
In resting cells, p105 is partially processed, generating p50. The exact mechanism of this limited processing event is not known, although co-translational processing events have been proposed (32,33). The unprocessed p105 functions as an inhibitor molecule and nonspecifically inhibits almost all NF-B subunits, including p50 (34 -36). An earlier study has shown that p105 retains itself, as well as other NF-B molecules, in the cytoplasm (37). The mechanism of inhibition is, however, still unclear. The p105 molecule, like the classical IB inhibitors, can also undergo complete degradation in response to signals and the sequential action of IB kinases, ubiquitin ligases, and the 26 S proteasome (38 -41). Both the death domain and the destruction box have been shown to be important for IB kinase phosphorylation (41,42). In contrast to p105, the mechanism of p100 processing is very different. In resting cells, most of p100 remains unprocessed, and in response to appropriate signals all of it gets processed into p52 (43,44).
In this study, we address the functional properties of p105 by asking the following questions. How does p105 regulate subcellular distribution of NF-B subunits? How does p105 nonspecifically inhibit other NF-B proteins? What is the relationship between p105⅐NF-B and IB␥⅐NF-B complexes? We find that although IB␥ resembles other IBs in exhibiting a 1:1 stoichiometry of binding to NF-B dimers, it has a unique specificity for NF-B dimers that contain at least one p50 subunit. This obligatory p50 subunit in IB␥⅐NF-B complexes is structurally equivalent to the Nterminal p50 segment of p105 in p105⅐NF-B complexes. Interestingly, although IB␥ fails to mask one NF-B NLS, it still retains NF-B in the cytoplasm. We observe that the death domain of p105 is necessary but not sufficient for this cytosolic retention. We suggest that p105 and IB␥ inhibitors may require an as yet unknown cellular factor(s) to sequester NF-B in the cytoplasm.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification from Escherichia coli-The cloning, expression, and purification of the NF-B subunits has been described previously (45,46). Full-length and truncated glutathione S-transferase-IB␥ were made by cloning into the pGEX-2T vector (Amersham Biosciences). The fusion protein was expressed in E. coli BL21 DE3 and purified by glutathione-agarose column chromatography following the manufacturer's protocol (Amersham Biosciences).
Native Polyacrylamide Gel Electrophoresis-Proteins and protein complexes were diluted in 10 mM Tris (pH 7.5), 200 mM NaCl, 4% glycerol, and 2 mM ␤-mercaptoethanol. The reactions were allowed to equilibrate at room temperature for 1 h. Native gel loading dye (50 mM Tris, pH 7.5, 0.1% bromphenol blue, 10% glycerol, and 1.25 mM ␤-mercaptoethanol) was then added to each sample. 10% native polyacryl-amide gels were prepared with 0.25ϫ Tris-borate-EDTA buffer. The samples were loaded on the gel and run in Tris-borate-EDTA buffer for 2 h at a constant current (3 mA). The protein bands were visualized by Coomassie staining.
Fluorescence Polarization Competition Assay-Fluorescence polarization competition assays were done as described previously (47). Briefly, varying concentrations of IB␥ were mixed with constant amounts of p50 homodimer pre-equilibrated with fluorescein-labeled DNA. The competition assay binding curves were analyzed for IC 50 values, defined as the concentration of IB␥ at 0.5 fractional occupancy.
Plasmids, Cell Culture, and Transfections-cDNAs encoding fulllength p50 and full-length/truncated p105 were cloned into a pcDNA vector (Invitrogen) containing either an N-terminal FLAG (p50 constructs) or an N-terminal HA (p105⅐IB␥ constructs) tag. HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, and antibiotics and transfected with plasmid DNA using the LipofectAMINE Plus reagent (Invitrogen). Protein expression was checked by Western blot.
Immunofluorescence-HeLa cells were grown on 8-well chamber slides (Lab Tek). The cells were transfected with a total of 0.2 g of plasmid DNA. After 24 h, the cells were washed with PBS and fixed with 3.7% formaldehyde in PBS for 10 min at room temperature. The cells were then permeabilized with 0.25% Nonidet P-40 in PBS for 1 min and blocked with 5 mg/ml bovine serum albumin in PBS containing 0.1% Tween 20 at room temperature for 30 min. Fluorescent detection was done by incubating the cells with monoclonal antibody 12CA5 (against HA), M2 (against FLAG), and H-286 (against p65) in PBS containing 5 mg/ml bovine serum albumin and 0.2% Nonidet P-40 at room temperature for 2 h. The cells were washed three times with buffer containing 0.2% Nonidet P-40 in PBS and incubated with fluoresceinated secondary antibody at room temperature for 1 h. Finally, the cells were washed three times with buffer containing 0.1% Tween 20 in PBS, and the slide was mounted with Vectashield (Vector Laboratories).
Immunoprecipitation-For immunoprecipitation from whole cell lysates, HeLa cells were harvested 24 h post-transfection and lysed with buffer containing 20 mM Tris (pH 7.5), 0.2 M NaCl, 1% Triton X-100, 1 mM EDTA, 2 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, and protease inhibitor mixture (Sigma). Immunoprecipitation was carried out by the protein A pull-down method. Briefly, 0.1-0.5 mg equivalent of protein from transfected cell extract was diluted with 300 l of lysis buffer. The appropriate antibodies were added to the extract and incubated overnight with protein A-Sepharose 4B beads (Sigma) at 4°C. The beads were pulled down by brief centrifugation and washed three times with the lysis buffer. The immunoprecipitates were then eluted from the beads with 2ϫ Laemmli buffer devoid of ␤-mercaptoethanol by heating at 95°C for 5 min. The bound proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane. The bound proteins were then identified by Western blot. In vitro immuno- precipitation was carried out with a similar protocol using purified E. coli proteins in place of whole cell extract.

Interactions between NF-B Dimers and IB␥-Unprocessed
p105 functions as an inhibitor of NF-B, although its mode of inhibition appears to be quite different from the prototypical IB proteins such as IB␣ and IB␤. As opposed to IB␣ and IB␤ that bind through p65, p105 inhibits all NF-B members nonspecifically. To understand the biochemical basis of NF-B inhibition by p105, we wanted to first test how the p105 C terminus (IB␥) binds to NF-B dimers. We have characterized the interactions between IB␥ and various NF-B dimers by protein-protein gel shift assays under native conditions. To simplify the assay we generated a truncated IB␥ containing only the ARD fused to a poly-His peptide. This construct will be referred to as ARD-IB␥⌬C. We observe that ARD-IB␥⌬C binds strongly to the p50⅐p50 homodimer and p50⅐p65 heterodimer but only weakly to the p65⅐p65 and c-Rel⅐c-Rel homodimer ( Fig. 2A). Although no stable complexes are formed between IB␥ and p65⅐p65 or c-Rel⅐c-Rel homodimers, we do not observe free IB␥ in these lanes. It is likely that these complexes are weaker and smear during electrophoresis. Based on these results we suggest that complexes between IB␥ and non-p50 containing NF-B dimers are highly unstable.
The fact that p105 binds to all NF-B proteins and IB␥ binds only to p50 dimers (any NF-B dimer that contains at least one p50 subunit) suggests that the required p50 subunit in IB␥⅐NF-B complexes is likely the N-terminal p50 segment of p105 in p105⅐NF-B complexes. Thus, the p105⅐p50 (or p105⅐p65) complex can be considered to be structurally equivalent to the IB␥⅐p50⅐p50 (or IB␥⅐p50⅐p65) complex (Fig. 2B).
The binding data presented above suggest that IB␥ prefers p50 containing NF-B dimers. However, these experiments do not reveal whether one or two molecules of IB␥ bind to one molecule of NF-B dimer. This is important considering that Bcl-3, which contains seven AR like I〉␥, has been proposed to bind to p50⅐p50 and p52⅐p52 homodimers in a 2:1 molar ratio (48 -50). To elucidate IB␥⅐NF-B binding stoichiometry, we have used two different IB␥ constructs, one is the full-length IB␥ expressed as a glutathione S-transferase fusion protein (FL-IB␥), and the other is the previously described ARD-IB␥⌬C. As seen in Fig. 2C (lanes 4 and 5), both of these proteins bind p50⅐p50 homodimer efficiently, and the resulting complex migrates to different positions in the native gel. In a reaction mixture containing both IB␥s and the p50 homodimer, binding of two molecules of IB␥ to the homodimer would result in a distinct, additional complex composed of these IB␥ molecules of different lengths and p50. However, when equivalent amounts of FL-IB␥, ARD-IB␥⌬C, and p50⅐p50 homodimer were mixed, only two distinct complexes, one corresponding to the FL-IB␥⅐p50⅐p50 complex and the other to the ARD-IB␥⌬C⅐p50⅐p50 complex, were seen (Fig. 2C, lane 6). Thus, we can conclude that one molecule of IB␥ binds to one p50⅐p50 homodimer.
The Role of the NF-B NLS in IB␥ Binding-The presence of one free NLS is a prime reason for the nucleocytoplasmic shuttling of the IB␣⅐NF-B complex. To elucidate the status of the NF-B NLS(s) in its complex with IB␥, p50⅐p50 homodimers of different lengths, both with and without the NLSs, were prepared, and their binding to ARD-IB␥⌬C was tested using native gel shift assays. Fig. 3A shows that all three of the p50⅐p50 dimers were capable of binding to ARD-IB␥⌬C, although the dimer with both NLSs deleted (p50 -350(ϪNLS)) seems to bind IB␥ relatively weakly (Fig. 3A, lanes 5-7). This suggests that IB␥ may not use the NLS sequence for p50 binding, or at least this region does not seem to contribute significantly. Similar binding patterns were also observed with FL-IB␥ (data not shown).
Co-immunoprecipitation experiments were performed to further confirm the presence of at least one free NLS in the IB␥⅐NF-B complex. If the p65 NLS in the IB␥⅐p50⅐p65 complex were free, then anti-p65 NLS antibody would interact with the p65 NLS. Here, the p50⅐p65 heterodimer was used because of the availability of the p65 NLS-specific monoclonal antibody. Free NF-B p50⅐p65 heterodimer and the IB␥⅐p50⅐p65 heterodimer complex were incubated with p65 NLS antibody. It was seen that the antibody was able to pull down NF-B in both the free and IB␥ complexed form (Fig. 3B, top panel). This demonstrates that the p65 NLS in the p50⅐p65 heterodimer is not protected by IB␥. As a parallel control, the IB␣⅐p50⅐p65 complex was also tested. The p65 NLS antibody was unable to pull down NF-B in the IB␣⅐p50⅐p65 complex (Fig. 3B, bottom panel). This corroborates with earlier results that have shown that the p65 NLS is masked in the IB␣⅐p50⅐p65 complex (26,27,52,53). Thus, these experiments show that at least one NF-B NLS remains unmasked in the IB␥⅐NF-B complex.
IB␥⅐NF-B and p105⅐NF-B Complexes Are Cytoplasmic-The above observation that at least one NF-B NLS is free in the IB␥⅐NF-B complex suggests that these complexes should localize to the nucleus. The subcellular distribution of the complex was tested by simultaneously transfecting HeLa cells with IB␥, p50, and p65. Immunostaining shows that free p50 and p65, which are most likely to be present as a heterodimer, are nuclear (Fig. 4A, left panel), but co-expression of IB␥ leads to cytoplasmic retention (Fig. 4A, right panel). Thus, although the IB␥⅐p50⅐p65 has at least one NLS free, the complex does not localize to the nucleus.
The subcellular distribution of p105 and IB␥ complexes with p50 was also tested. Immunofluorescence studies show both these complexes to be cytoplasmic (Fig. 4B). Free p50, as expected, is localized in the nucleus (data not shown).
To further test the ability of these complexes to be retained in the cytoplasm, a mutant p50 with an additional NLS (p50-NLS) was made. Cells co-transfected with p50-NLS and p105 or IB␥ reveal that the complexes are retained in the cytoplasm (Fig. 4C, left panels). To test whether the p105⅐p50-NLS complex shuttles between the nucleus and cytoplasm, HeLa cells were treated with the nuclear export inhibitor leptomycin B (LMB), and the subsequent changes in localization were mon-itored. LMB treatment does not appear to alter the localization of the p105⅐p50-NLS or the IB␥⅐p50-NLS complexes (Fig. 4C, right panels), implying that these complexes do not shuttle. This result is in contrast to that observed for IB␣⅐NF-B complexes where addition of the inhibitor confines both proteins within the nucleus (26). Thus, the p105⅐NF-B complexes seem to be similar to the IB␤⅐NF-B complexes, which are primarily cytoplasmic. Free p50-NLS, as expected, is localized within the nucleus (data not shown).
The Death Domain Is Necessary for Cytoplasmic Retention of NF-B-To assess the possible role of the death domain and the C-terminal tail containing the destruction box in the cytoplasmic retention of NF-Bs, deletion mutants of both p105 and IB␥ were constructed. HeLa cells were co-transfected with p50 and various truncated forms of p105 or IB␥. Subcellular localization of these complexes was monitored by immunofluorescence. As mentioned in the previous results section, the p105⅐p50 complex is cytoplasmic (Fig. 5A, left panel). Removal of the last ϳ100 residues, which includes the destruction box, p105⌬DB, did not alter the localization of the complex (Fig. 5A, right panel). However, deletion of the C-terminal 171 residues (encompassing the death domain), p105⌬C, localizes the complex exclusively in the nucleus (Fig.  5A, middle panel). These results suggest that the death domain of p105⅐IB␥ plays a role in the retention of NF-B complexes within the cytoplasm. p50 is again retained in the cytoplasm when co-expressed with a truncated p105, p105⌬N (Fig. 1), where the N-terminal, DNA-binding immunoglobulin domain (residues 1-244) is deleted (Fig. 5B, left panel). However, when both the N-and C-terminal domains (p105⌬N⌬C) are deleted, these proteins localize to the nucleus (Fig. 5B, right panel).
Similar results were obtained with various deletion mutants of IB␥. Co-expression of full-length IB␥ and p50 lead to cytoplasmic retention of the complex (Fig. 5C, left panel). Deletion of the C terminus, IB␥⌬C, leads to nuclear localization of both proteins (Fig. 5C, right panel). Also, a shorter construct of IB␥ comprising the ARD and all residues downstream, ARD-IB␥ (Fig. 1) and the corresponding C terminally deleted form, ARD-IB␥⌬C, showed a similar retention pattern (Fig.  5D).
It is known that NF-B dimers are nuclear proteins, and it is only in complex with IB that they are retained in the cytoplasm. As described above, co-expression of p50 and death domain-deleted p105 or IB␥ leads to nuclear localization of both components. To test whether these proteins localize independently or are present as a complex in the nucleus, coimmunoprecipitation experiments were done. FLAG antibody (M2) was used to immunoprecipitate FLAG-tagged p50 from whole cell lysates that had been co-transfected with FLAG-p50 and HA-p105 or HA-IB␥. Subsequent Western blotting was done using the HA antibody. The results clearly show that FIG. 5. Death domain is necessary for cytoplasmic retention. A, co-transfection of full-length p105 with p50 leads to cytoplasmic localization of both proteins (left panel). Co-transfection of death domain and destruction box truncated p105 (p105⌬C) with p50 leads to nuclear accumulation of both proteins (middle panel). Co-transfection of destruction box truncated p105 (p105⌬DB) with p50 leads to cytoplasmic accumulation of both proteins (right panel). B, co-transfection of N-terminally truncated p105, p105⌬N, with p50 leads to cytoplasmic localization of both proteins (left panel). Co-transfection of death domain and destruction box truncated p105⌬N (p105⌬N⌬C) with p50 leads to nuclear accumulation of both proteins (right panel). C, co-transfection of full-length IB␥ with p50 leads to cytoplasmic localization of both proteins (left panel). Co-transfection of death domain and destruction box truncated IB␥ (IB␥⌬C) with p50 leads to nuclear accumulation of both proteins (right panel). D, co-transfection of ARD-IB␥ with p50 leads to cytoplasmic localization of both proteins (left panel). Co-transfection of death domain and destruction box truncated ARD-IB␥ (ARD-IB␥⌬C) with p50 leads to nuclear accumulation of both proteins (right panel). E, co-immunoprecipitation of the C-terminally truncated p105 and IB␥ with p50. Western blot (WB) showing the binding of HA-tagged, death domain-deleted, and destruction box-deleted p105 (and IB␥) to FLAG-tagged p50. HeLa cells were co-transfected with p50 and various truncated forms of p105 or IB␥s. The p105⅐p50 or the IB␥⅐p50 complex was co-immunoprecipitated (IP) by incubating with anti-FLAG antibody. Western blotting was done to detect p105 or IB␥ using anti-HA antibody. Control experiments were done with cells transfected with only p105 or IB␥s in the absence of p50. All of the C-terminally truncated p105 and IB␥ seem to form a stable complex with p50. these proteins form a complex because the FLAG antibody was able to pull down both p50 and p105 (or IB␥) (Fig. 5E). In all, the above experiments suggest that the death domain of p105⅐IB␥ is necessary for cytoplasmic retention.
The Death Domain Is Not Sufficient for Cytoplasmic Retention-Next, we wanted to test whether the death domain was the exclusive retention signal in these IB molecules. Fulllength I〉␥ when expressed in HeLa cells exhibits a nucleocytoplasmic distribution (Fig. 6A, left panel). The death domaintruncated IB␥⌬C, as expected, was localized within the nucleus (Fig. 6A, right panel). Both ARD-IB␥ and ARD-IB␥⌬C, by themselves, were present in the nucleus (Fig. 6B). Thus, the presence of the death domain is not sufficient for cytoplasmic localization.
To further test the role of the death domain, two more constructs were made. One containing only the death domain, p105 (800 -887), and the other with all the segments further downstream, p105 (800 -971). Both of these proteins localized within the nucleus (Fig. 6C). This suggests that neither the death domain by itself nor its presence in the IB␥ molecule provides the retention signal.
The Death Domain Does Not Participate in NF-B Binding-Because the death domain of p105⅐IB␥ seems to be important for cytoplasmic localization of NF-B, a possible model of the complex could be one in which the death domain binds and masks the exposed NLS of NF-B, thereby preventing nuclear import. If this were true, then full-length IB␥ must interact with the p50⅐p50 homodimer with higher affinity than the death domain-truncated IB␥. Although the native gel shift assays showed that the death domain was not essential for NF-B binding, these experiments were of a qualitative nature. Fluorescence polarization competition assays were done to determine the binding affinity of NF-B for both full-length and death domain-truncated IB␥. In a solution-based competition assay, fixed amounts of fluorescein-labeled B DNA bound to the p50⅐p50 dimer was incubated with increasing amounts of IB␥. In this experiment, the IB␥-dependent dissociation of the p50⅐p50 dimer from DNA is accompanied by a distinct change in fluorescence polarization. The DNA binding inhibition constant at equilibrium was then calculated. As shown in Fig. 7, both IB␥s inhibit NF-B DNA binding to a similar extent. Thus, the death domain does not seem to play a major role in NF-B binding to IB␥.
Limited Processing of p105: a Possible Model of the Cytoplasmic Complex-In addition to testing the localization of p50 complexes with the full-length and truncated forms of p105, we have also expressed just p105 and its truncated forms in HeLa cells and tested their individual subcellular distribution. p105 is present exclusively in the cytoplasm (Fig. 8A, left panel), whereas the C-terminally truncated p105, p105⌬C, was exclusively nuclear (Fig. 8A, right panel). The N-terminally truncated p105, p105⌬N, also localized to the cytoplasm (Fig. 8B, left panel). The corresponding Cterminal deleted p105, p105⌬N⌬C, was nuclear (Fig. 8B,  right panel). Localization patterns of p105 and its derivatives are thus identical to that observed previously for the p105⅐p50 complexes in HeLa cells, expressing both p105 (or its derivatives) and p50.
p105 is known to undergo limited processing to generate p50 molecules. To test whether the other truncated forms of p105 were also capable of similar processing, Western analysis was done using HA monoclonal antibody. We observe that like full-length p105, its truncated forms were also processed in a limited manner within the cell. The appearance of an additional band of lower molecular weight indicates a processed band, derived from the C-terminal processing of the intact protein. The lengths of the processed products vary according to the truncations of p105. Cells transfected with the C-terminally truncated form, p105⌬C, show a processed band corre-

FIG. 7. Death domain is not involved in NF-B binding.
The fluorescence polarization competition assays were performed to determine the binding affinities of the IB␥⅐p50 complexes. The affinity of full-length IB␥⅐p50 was found to be 38 nM (green) and that of the C-terminally deleted IB␥⌬C⅐p50 was found to be 45 nM (red). sponding to full-length p50, and cells transfected with the smaller, N-terminally truncated form, p105⌬N and p105⌬N⌬C, generate a shorter p50 (Fig. 8C). On the other hand, in cells transfected with both truncated p105 and p50, no processed p50 band is detected. This observation corroborates with that seen earlier for full-length p105 and p50 (41). These results lead us to suggest that p105 is first processed in a limited manner and that the resulting complex formed between the unprocessed p105 and the processed p50 is identical to the one in which both proteins are expressed within the same cell.
A minimal complex formed between p105 and p50 seems to require just the dimerization domain of p50. Therefore, the smallest p105 derivative, p105⌬N⌬C, can be partially processed into a truncated p50 comprising of only the dimerization domain (p50⌬N), and this can associate with unprocessed p105⌬N⌬C. Such a complex would be equivalent to the complex between the p50 dimerization domains and the ARD-IB␥⌬C. Indeed, native gel shift assays show that the dimerization domain of p50 (p50⌬N-376 and p50⌬N-363) can form a stable complex with ARD-IB␥⌬C (Fig. 8D). However, the p105⌬N⌬C⅐p50⌬N complex is not cytoplasmic because the death domain is absent. In contrast, the p105⌬N⅐p50⌬N complex is cytoplasmic.
These results indicate that in addition to the death domain, p105⅐NF-B or the equivalent IB␥⅐NF-B complex formation is necessary for cytoplasmic retention. In the retention process, only the N-terminal domain (residues 1-244) and the C-terminal region (residues 885-971) of p105 are dispensable. DISCUSSION Earlier studies using competition assays have shown that IB␥ preferentially inhibits p50⅐50 homodimer (18,31,51). In the present study we have performed direct binding assays to demonstrate that IB␥ interacts stably with both p50⅐p65 and p50⅐p50 dimers but not with p65⅐p65 or c-Rel⅐c-Rel dimers. This suggests that IB␥ may have a preference for p50 containing NF-B dimers. We and others have previously shown that IB␣ interacts strongly with p65 containing dimers (23,24,26). The x-ray structure of IB␣⅐p50⅐p65 heterodimer showed that the p65 subunit makes extensive contacts with IB␣ (52,53). Thus, a central theme seems to govern all IB⅐NF-B binding wherein the stability of the interaction is dictated by the association of an IB molecule with one specific NF-B subunit. The second subunit, which we refer to as the nonspecific subunit, associates with the specific subunit through its dimerization domain. This is illustrated in Fig. 9A. The nature of the specific subunit is directly correlated to the nature of the binding IB molecule. For IB␣, p65 serves as the specific subunit. IB␤ also exhibits similar dimer specificity, requiring p65 as the primary binding partner. On the other hand, in IB␥, the specific subunit appears to be p50. In p105⅐NF-B complexes, by the very nature of the molecule, the docking occurs intramolecularly with the N-terminal p50 arm serving as the specific subunit. It is possible that in IB⅐NF-B complexes the nonspecific subunit can be exchanged with other nonspecific subunits, and this may dictate the cellular pool of active NF-Bs.
Sequestration of NF-B in the cytoplasm can be considered to be one of the most simplistic ways by which IB molecules exhibit their inhibitory function. They can do this by blocking FIG. 8. Subcellular localization of full-length and N-terminally truncated p105. A, transfection of full-length p105 leads to cytoplasmic localization of the protein (left panel). Transfection of death domain-and destruction box-truncated p105 (p105⌬C) leads to nuclear accumulation of the protein (right panel). B, transfection of N-terminally truncated p105, p105⌬N, leads to cytoplasmic localization of the protein (left panel). Transfection of death domain and destruction box truncated p105 (p105⌬N⌬C) leads to nuclear accumulation of the protein (right panel). C, processing of p105 in the presence and absence of co-expressed p50. Western blot showing the processing of various p105 mutants. HeLa cells were transfected with C-terminally (p105⌬C), Nterminally (p105⌬N), and both C-and N-terminally (p105⌬N⌬C) truncated p105 in the presence and absence of p50. As seen, all three mutants undergo processing (lanes 1-3), and this is abrogated in the presence of co-expressed p50 (lanes 4 -6). D, N-terminally deleted p50 binding by ARD-IB␥⌬C. Native gel mobility shift assay of ARD-  IB␥⌬Cϩp50⌬N-363. p50⌬N-376  and p50⌬N-363 represent p50 (245-376) and p50 (245-363), respectively. The free and complexed proteins are indicated by arrows.
the NLS of NF-B dimers. However, both IB␣ and IB␤, upon binding to the specific NF-B dimers, mask only the NLS of the specific NF-B subunit. As expected therefore, IB␣⅐NF-B complexes are not truly cytoplasmic but dynamic nucleocytoplasmic (23)(24)(25)(26). We observe that in the IB␥⅐p50⅐p65 complex the p65 NLS is free. Although no structural evidence exists to confirm that the NLS of the specific subunit is completely masked, two lines of evidence indicate this to be most likely. First, we show that when both NLSs are removed the affinity of the p50⅐p50 homodimer for IB␥ appears to be reduced, and second, we show that the NLS of the nonspecific p65 subunit of the IB␥⅐p50⅐p65 complex is free. Therefore, reduced binding affinity of p50⅐p50 homodimer for IB␥ is likely due to the removal of the NLS of the p50 subunit that makes specific contacts with IB␥.
The domain architecture of the C-terminal inhibitory domain of p105 is different from that of IB␣ and IB␤. p105 does not contain a PEST sequence in its C-terminal tail as in IB␣ and IB␤, but it does contain a death domain. The PEST sequence in IB␣ and IB␤ is important for NF-B binding. In case of IB␥, we observe that the death domain does not play any role in NF-B binding, suggesting that this domain is not involved in masking the exposed NLS of the nonspecific NF-B subunit in p105⅐NF-B or IB␥⅐NF-B complexes. The death domain,  (p105⅐NF-B) where the specific subunit is p50. B, a possible model of the cytoplasmic p105⅐NF-B complex. Transfected p105 is capable of undergoing limited processing, which is coupled to a dimerization event leading to the formation of a p105⅐p50 complex. Co-expression of both p105 and p50 can lead to a similar complex formation without processing. The NLS in this complex seems to be exposed, and yet it is retained in the cytoplasm. A possible mode of sequestration could be binding to another cytoplasmic protein(s) through the scaffold formed because of the association of p105 and p50 and the death domain of p105. however, seems critical for cytoplasmic retention. When this domain is removed from p105⅐IB␥, the subcellular localization of the truncated molecules in their free form or their complexes with NF-B change completely, from cytoplasmic to nuclear. This observation is consistent with earlier studies showing that removal of C-terminal 191 residues alters the localization from the cytoplasm to the nucleus (37). However, although this clearly suggested the role of the C-terminal domain of p105, this large truncation also removed most of the last AR. Therefore, it was not clear whether removal of part of the ARD contributed to altered localization of p105. Interestingly, the death domain, by itself, does not contain any retention signal because we see that the free death domain localizes to the nucleus. Full-length or N-terminally truncated IB␥, both of which contain the death domain, are also primarily nuclear.
We observe that free p105, as in the p105⅐p50 complex, is also cytoplasmic. However, p105 does not exist in its free form in cells. The molecule is partly processed generating p50, and this processed p50 forms a dimer with the precursor p105. p105⅐p50 complex formation is necessary for cytoplasmic sequestration of both p50 and p105. Our results show that limited processing does not produce an exactly 1:1 molar ratio of p105. It is thus possible that excess unprocessed p105 associates with other NF-Bs such as p65 and c-Rel, and the p105⅐NF-B complex is retained in the cytoplasm. Cellular localization of truncated p105 is dependent upon the presence or absence of the death domain. Our results suggest that the three-dimensional scaffold formed because of the association of p105 and p50 (or another NF-B subunit), in addition to the death domain, is required for the cytoplasmic retention of the complex. Association between p105 and p50 (or any other NF-B subunit) requires only the dimerization domain of p50 and the ARD of IB␥. However, the NLS of the nonspecific subunit is still exposed to the solvent. How is the complex then retained in the cytoplasm? We do not yet know the precise mechanism of cytoplasmic sequestration of this complex.
Earlier work in our laboratory has shown that IB␤⅐NF-B complexes are cytoplasmic, although even here, as in IB␣⅐NF-B complexes, one NLS is mostly solvent-exposed. Recent studies lead us to believe that IB␤⅐NF-B complexes interact with additional cellular proteins to mask the exposed NLS, thereby sequestering them in the cytoplasm. 2 It is possible that p105 employs a similar mechanism, but not necessarily the exact same mechanism. Thus, a bridging factor, which requires the death domain and the NF-B scaffold, could be involved in masking the NLS. The death domain has been shown to confer various homotypic and heterotypic interactions in a large number of signaling pathways (54). It may thus play a major role in binding to the bridging factor. Consistent with this model, an earlier study demonstrated the inaccessibility or the reduced accessibility of the Cterminal tail in p105 by using antibodies raised against the C-terminal (37). This study also showed that the NLS of p105 is masked. Fig. 9B thus summarizes a likely sequence of events based on the above results. Expression of p105 or both p105 and p50 results in a p105⅐p50 complex that is retained in the cytoplasm. As noted earlier, the N-terminal domains and the last ϳ100 residues of p105 do not seem to be a part of the retention signal. Thus, a plausible model could be one in which a composite surface formed by the coming together of the relevant regions of the protein complex can interact with a putative, cytoplasmic protein that leads to its docking. This could then explain the inability of the p105⅐NF-B or the IB␥⅐NF-B complex to move to the nucleus even with an additional NLS. We conclude that the IB family proteins employ diverse mechanisms for cytoplasmic sequestration of NF-B dimers. Elucidating each of these mechanisms in detail is imperative for understanding their cellular function and participation in unique signal transduction pathways.