The N-Terminal Domain of IκBα Masks the Nuclear Localization Signal(s) of p50 and c-Rel Homodimers

Abstract

Members of the Rel/NF-κB family of transcription factors are related to each other over a region of about 300 amino acids called the Rel Homology Domain (RHD), which governs DNA binding, dimerization, and binding to inhibitor. At the C-terminal end of the RHD, each protein has a nuclear localization signal (NLS). The crystal structures of the p50 and RelA family members show that the RHD consists of two regions: an N-terminal section which contains some of the DNA contacts and a C-terminal section which contains the remaining DNA contacts and controls dimerization. In unstimulated cells, the homo- or heterodimeric Rel/NF-κB proteins are cytoplasmic by virtue of binding to an inhibitor protein (IκB) which somehow masks the NLS of each member of the dimer. The IκB proteins consist of an ankyrin-repeat-containing domain that is required for binding to dimers and N- and C-terminal domains that are dispensable for binding to most dimers. In this study, we examined the interaction between IκBα and Rel family homodimers by mutational analysis. We show that (i) the dimerization regions of p50, RelA, and c-Rel are sufficient for binding to IκBα, (ii) the NLSs of RelA and c-Rel are not required for binding to IκBα but do stabilize the interaction, (iii) the NLS of p50 is required for binding to IκBα, (iv) only certain residues within the p50 NLS are required for binding, and (v) in a p50-IκBα complex or a c-Rel-IκBα complex, the N terminus of IκBα either directly or indirectly masks one or both of the dimer NLSs.

Transcription factors of the Rel/NF-κB family are present in most or all mammalian and avian cells and influence the expression of many genes (for a review, see references 1 and 19). These factors are related to each other over a region of about 300 amino acids called the Rel Homology Domain (RHD), which governs DNA binding and dimerization. Although each protein has a nuclear localization signal (NLS), the various homo- and heterodimers are cytoplasmic by virtue of an inhibitor protein whose binding masks the NLSs. Several different inhibitors (called IκBs) are known, and they constitute a family of related proteins. Each IκB contains five or six so-called ankyrin repeats, motifs of about 33 amino acids which are abundantly represented in the protein ankyrin. In ankyrin, these repeats apparently function in groups of about six (18) and constitute a protein-protein interaction domain. In addition, the inhibitors contain a C-terminal region which tends to be highly acidic and to resemble PEST sequences associated with rapid protein turnover (22). The IκB N-terminal region contains two serine residues which, upon phosphorylation, trigger ubiquitination and proteasomal degradation of the inhibitor, thus freeing the dimer to enter the nucleus (4, 5, 7, 26–29).

While several crystal structures of Rel family members are known (6, 6a, 10, 20), none is known for the IκB family. Nevertheless, studies with various mutants of the Rel family and/or IκB proteins have led to some general conclusions about their interaction. First, a Rel family dimer binds a single IκB molecule (8, 12, 14). Second, at least some of the contacts with IκB occur in the dimerization domains of the Rel proteins. Based on the crystal structure, each monomer of p50 and RelA consists of two separate domains connected by a short hinge region. The N-terminal domain contains some (but not all) of the residues that contact DNA, while the C-terminal domain contains the remaining DNA contacts and governs dimerization. Partial deletion of the N-terminal domain of RelA did not abolish binding of IκBα (9). Similar results were obtained with Cactus and N-terminal deletion mutants of Dorsal (Drosophila relatives of IκB and Rel proteins, respectively) (11, 25). In addition, mutation of two residues in the dimerization domain of Dorsal prevented interaction with Cactus but did not affect DNA binding or dimerization (16). These studies suggest contact between IκB and the C-terminal part of the RHD. The third general conclusion is that the dimer NLSs, which were not resolved in the crystal structures but are immediately downstream of the dimerization domain in the primary structure, are somehow involved in binding to IκB. While in at least some cases an intact NLS may not be required for binding to IκB (11, 23), several studies have shown that altering the NLS can abolish binding (2, 9, 15, 25). Involvement of the NLS would correlate nicely with the fact that the dimer NLSs are masked in a dimer-IκB complex. The final generalization is that most of the IκB contacts occur within the ankyrin repeats. Mutations within the ankyrin domain disrupt binding, while deletion of the N terminus and/or the C terminus usually does not (8, 12, 13, 17, 30).

In this study, we examined in more detail the interaction between IκBα and Rel family homodimers. We show that (i) the dimerization domains of p50, RelA, and c-Rel are sufficient for binding to IκBα both in vivo and in vitro, (ii) the NLSs of RelA and c-Rel are not required for binding to IκBα but do stabilize the interaction, (iii) the NLS of p50 is required for binding to IκBα, (iv) only certain residues within the p50 NLS are required for binding, and (v) in a p50-IκBα complex or a c-Rel-IκBα complex, the N terminus of IκBα either directly or indirectly masks one or both of the dimer NLSs.

MATERIALS AND METHODS

Plasmids and mutagenesis.

Human IκBα cDNA, a gift of Al Baldwin, was cloned into the plasmid Rc/CMV (Invitrogen). Human IκBα mutants from which residues 2 to 30 or 2 to 53 had been deleted were gifts of Simon Whiteside; an IκBα mutant from which residues 44 to 50 had been deleted was a gift of Masashi Muroi. Wild-type p50 (amino acids 1 to 399 of human p105) in Rc/CMV was a gift of Alain Israël. C-terminal truncations of c-Rel and RelA were created to eliminate their transactivation domains. The resulting c-RelΔC (residues 1 to 361 of human c-Rel) has been described earlier (24) and was used both in Rc/CMV and in a modified Bluescript vector containing the Rous sarcoma virus long terminal repeat (24). The plasmid RelAΔC was created from the plasmid pCMVINp65 (a gift of Craig Rosen) by PCR to amplify the cDNA, while the coding region after residue 337 was truncated. The codon for Ser 338 was converted to a TAA stop codon. At the same time, 5′ BamHI and 3′ XhoI sites were engineered up- and downstream of the coding sequence. The resulting PCR product was cleaved with BamHI and XhoI and cloned into the corresponding sites of the CMV-based expression vector pcDNA3 (Invitrogen).

To create N-terminal deletions of p50, c-Rel, and RelA, oligonucleotide-directed mutagenesis was used to insert ClaI sites directly following the N-terminal methionine codon and directly preceding the codon encoding p50 Pro 246, c-Rel Pro 189, or RelA Pro 181. In the p50 crystal structure (10, 20), this proline is within the proposed hinge region that separates the N- and C-terminal domains. The resulting DNAs were digested with ClaI and religated to excise the N-terminal domain. The mutants had the following N-terminal sequences: p50ΔN, MIDPNASNLKIV…(residues 2 to 245 deleted); c-RelΔCΔN, MIDPNTAELRIC…(residues 2 to 180 deleted); RelAΔCΔN, MIDPNTAELKIC (residues 2 to 188 deleted). In each case, there is a two-amino-acid insertion (ID) due to the remaining ClaI site.

With one exception, all other mutants were created by oligonucleotide-directed mutagenesis with the Bio-Rad Muta-Gene Phagemid In Vitro Mutagenesis kit, version 2. Deletion of the c-Rel NLS (to give the plasmid c-RelΔCΔNLS) was performed with the Transformer Mutagenesis kit from Clontech. The sequences of all mutagenic oligonucleotides are available upon request. Dideoxy chain termination sequencing with the Sequenase 2.0 kit (U.S. Biochemicals) was performed on all mutant plasmids to confirm mutagenesis and to ensure that the mutant plasmid DNAs were free from contamination by wild-type plasmids.

Transfection.

Human kidney 293 cells were seeded at 10⁶/6-cm-diameter dish. Twenty-four hours later, they were transfected by the calcium phosphate method.

Immunofluorescence.

Our procedure has been described previously (3). Briefly, transfected 293 cells were seeded onto collagen-coated glass coverslips 24 h after transfection. After an additional 16 h, the cells were fixed in 2% paraformaldehyde, quenched in phosphate-buffered saline containing 50 mM glycine, rinsed, permeabilized in Tris-buffered saline (pH 7.4) containing 0.1% Nonidet P-40 (NP-40) and 0.5 M NaCl, blocked with goat serum and chicken serum albumin, and incubated with primary antibody at 1:40 or 1:80 for 1 h at room temperature. The secondary antibody was goat anti-rabbit immunoglobulin G (IgG) conjugated with fluorescein isothiocyanate (Kirkegaard and Perry Laboratories). In some cases, treatment with primary antibody was followed by incubation with biotinylated goat anti-rabbit IgG (Vector Laboratories) and then with fluorescein-conjugated streptavidin (Immunotech). After a final washing in phosphate-buffered saline, coverslips were dried and mounted on slides with Vectashield (Vector Laboratories).

Antisera.

Rabbit antisera raised against synthetic peptides were employed in all experiments. Most of them have been described previously (8, 21, 24). Antisera that recognize p50 (and boundaries of the peptides in the human protein) are as follows: no. 1141 (residues 2 to 15), no. 1157 (residues 339 to 357), and no. 1613 (residues 354 to 374). Antisera that recognize RelA are no. 1207 (residues 2 to 17), no. 1774 (residues 314 to 329), and no. 1226 (residues 537 to 550). The c-Rel antiserum was no. 1136 (residues 305 to 320). Antisera that recognize IκBα are no. 1309 (residues 2 to 15) and no. 1258 (residues 301 to 317).

Metabolic labeling and immunoprecipitation.

Twenty-four hours after transfection, cells were grown in media containing [³⁵S]methionine and [³⁵S]cysteine (each at 50 μCi/ml; 1,000 Ci/mmol; Amersham) for 1 h. For whole-cell extracts, cells were lysed and immunoprecipitated in ELB buffer (50 mM HEPES [pH 7.0], 250 mM NaCl, 5 mM EDTA, 0.1% NP-40) containing protease inhibitors. For cytoplasmic extracts, cells were lysed in HB buffer (25 mM Tris [pH 7.4], 5 mM KCl, 1 mM MgCl₂, 0.05% NP-40) and nuclei and insoluble material were removed by centrifugation at 10,000 × g for 5 min. For immunoprecipitation of the cytoplasmic extracts, NaCl and Triton X-100 were added to final concentrations of 0.1 M and 0.5%, respectively. Precipitates were collected on protein A-Sepharose (Pharmacia) and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Usually 5% of the total lysate from a 6-cm-diameter dish was used per lane, giving a strong signal on X-ray film after overnight exposure.

Immunoprecipitation and immunoblotting.

Transfected cells were lysed and immunoprecipitated in ELB buffer, as described above, but without prior metabolic labeling. Precipitates were subjected to SDS-PAGE, and the gel was blotted onto an Immobilon membrane (Millipore). The immunoblot was incubated first with primary antiserum at a 1:1,000 dilution and then with peroxidase-conjugated goat anti-rabbit IgG (Boehringer Mannheim) at a 1:10,000 dilution and developed by the Amersham enhanced chemiluminescence system.

EMSA.

Transfected or cotransfected cells were lysed in ELB buffer and analyzed as previously described (8). The ³²P-oligonucleotide probe contained the κB binding site from the interleukin-6 promoter. Extracts and probe, with or without antisera, were mixed in electrophoretic mobility shift assay (EMSA) buffer: 10 mM HEPES (pH 7.5), 80 mM KCl, 1 mM EDTA, 1 mM EGTA, 6% glycerol, 0.05 μg of salmon sperm DNA per μl, and 0.05 μg of poly(d[IC]) per μl.

RESULTS

The N-terminal portion of the RHD is not required for IκBα binding.

With the X-ray crystal structure of the p50 homodimer as a guide (10, 20), mutations were introduced in the coding sequences of the human p50, human RelA, and human c-Rel cDNAs, enabling us to delete the entire N-terminal section of the RHD. We also deleted the transactivation domains of RelA and c-Rel in order to avoid induction of endogenous IκBα upon transfection. The resulting constructs (RelAΔCΔN and c-RelΔCΔN) consisted almost entirely of the dimerization region of the RHD (Fig. 1). These truncated proteins were then expressed in human 293 cells either alone or in combination with the human IκBα protein. Subcellular localization of the overexpressed proteins was assayed by indirect immunofluorescence with antisera specific for the NF-κB proteins. As shown in Fig. 2A, the N-terminally truncated p50 construct (p50ΔN) was strictly nuclear (panel a). When this p50 mutant was coexpressed with IκBα in excess, however, it was cytoplasmic (panel b), indicating that IκBα bound to the mutant and retained it in the cytoplasm. Similarly, the comparable RelA mutant protein (RelAΔCΔN) was nuclear when expressed alone (panel c) but was retained in the cytoplasm when it was coexpressed with IκBα (panel d). The same results were obtained with c-RelΔCΔN (panels e and f). Based on these results, we conclude that there are no sites in the N-terminal domains of these NF-κB proteins that are necessary for IκBα binding in vivo or retention of the NF-κB homodimers in the cytoplasm.

FIG. 1.

Deletion mutants of p50, RelA, and c-Rel. Within each of the three sets, the wild-type protein is diagrammed on top. N- and C-terminal residue numbers in the ΔN and ΔCΔN mutants are indicated. Numbers for the ΔNLS mutants indicate the deleted residues. A summary of results presented in the text is shown at the right.

FIG. 2.

Rel family proteins lacking the N-terminal domain are able to bind IκBα. (A) 293 cells were transfected with N-terminal deletion mutants of p50, RelA, or c-Rel or cotransfected with IκBα in excess as indicated for each panel. Localization of the mutant proteins was assayed by immunofluorescence with antisera against p50 (no. 1613), RelA (no. 1774), or c-Rel (no. 1136). The targets of the antisera in the cotransfection experiments (b, d, and f) are underlined. (B) 293 cells were transfected, cotransfected, or mock transfected, as indicated. Whole-cell lysates were immunoprecipitated with anti-IκBα (no. 1309) (lanes 1, 2, and 4), preimmune serum (lane 3) or antiserum raised against the N terminus of RelA (no. 1207, RelA[N]) (lane 5), a peptide located within RelAΔCΔN (no. 1774, RelA[I]) (lane 6), the C terminus of wild-type RelA (no. 1226, RelA[C]) (lane 7), or a peptide located within c-Rel ΔCΔN (no. 1136, c-Rel[I]) (lanes 8 to 11). Immunoprecipitates were fractionated by SDS-PAGE, and immunoblots were analyzed with anti-p50 (no. 1157) (lanes 1 and 2) or anti-IκBα (no. 1309) (lanes 3 to 11). IκBα did not precipitate with antisera raised against the N or C terminus of wild-type RelA (lanes 5 and 7), proving that the precipitated IκBα in lane 6 was bound to RelAΔCΔN and not to endogenous full-length RelA. Ab for IP and Western, antibody used for immunoprecipitation and immunoblotting, respectively.

To investigate the stability of the mutant NF-κB–IκBα complexes, we tested whether they remain intact in vitro with buffer containing 0.2 M NaCl and 1% nonionic detergent. After lysis of the transiently transfected cells in this buffer, the extracts were immunoprecipitated with antiserum raised against one member of the complex and precipitates were assayed for the other member by immunoblotting. As shown in Fig. 2B, IκBα coprecipitated with all three N-terminally truncated proteins (lanes 2, 6, and 11). Thus, complexes containing these mutants and IκBα are sufficiently stable to survive lysis of the cell under conditions of moderate salt and nonionic detergent.

The NLS of p50, but not that of RelA or c-Rel, is required for IκBα interaction.

Having shown that the C-terminal portions of the RHDs of p50, RelA, and c-Rel are sufficient for binding to IκBα, we looked for critical residues within that region. Several groups have reported that the NLSs of Rel family proteins are required for IκBα binding (2, 9, 15, 25), while others have found that they are not (11, 23). To investigate this problem, oligonucleotide-directed mutagenesis was used to delete the NLSs of p50, RelA, and c-Rel. These mutant proteins were then assayed in vivo and in vitro for their subcellular localizations and abilities to bind IκBα.

Overexpression of the p50 NLS deletion mutant (p50ΔNLS) in human 293 cells yielded a cytoplasmic protein (Fig. 3A, panel b), indicating that, as expected, the residues deleted were important for the nuclear localization of p50. To test for interaction of the ΔNLS mutants with IκBα, we assayed the localization of IκBα in the presence of excess dimer. When IκBα is expressed in the absence of proteins to which it can bind, it is distributed throughout the cell (panel c). In contrast, when cotransfected with excess protein to which it can bind, such as wild-type p50, IκBα is cytoplasmic (panel d). The result of cotransfection with p50ΔNLS was that, although p50ΔNLS was clearly expressed (panel f), it did not result in cytoplasmic retention of IκBα (panel e), indicating that the two proteins are unable to bind to each other. The ability of IκBα to bind p50ΔNLS was also tested in vitro by immunoprecipitation. In lysates of cells cotransfected with p50 and IκBα, antiserum to p50 coprecipitated IκBα (Fig. 3B, lane 1) and antiserum to IκBα coprecipitated p50 (lane 2). However, when cells were cotransfected with p50ΔNLS and IκBα, anti-p50 precipitated only p50ΔNLS (lane 3) and anti-IκBα precipitated only IκBα (lane 4). Therefore, in agreement with others (2), we conclude that the NLS of p50 is required for IκBα interaction both in vivo and in vitro. Inability to bind IκBα could be because p50ΔNLS failed to dimerize. However, p50ΔNLS showed no impairment in ability to bind DNA (data not shown), indicating that dimerization was not significantly affected by the deletion. This result suggests that binding of p50 to IκBα involves direct contact of IκBα with the p50 NLS.

FIG. 3.

IκBα binds to RelA- and c-Rel mutants lacking an NLS but not to NLS⁻ p50. (A) 293 cells were transfected or cotransfected as indicated, and localization of the expressed proteins was assayed by immunofluorescence. For cotransfected cells, the targets of the antisera are underlined. Nonunderlined partners were present in excess in panels d, e, i, j, and m. (B) 293 cells were transfected, cotransfected, or mock transfected, as indicated. Twenty-four hours later, the cells were incubated in medium containing ³⁵S-amino acids and whole-cell lysates were immunoprecipitated with anti-p50 (no. 1141) (lanes 1 and 3), anti-IκBα (no. 1309) (lanes 2 and 4), or anti-c-Rel (no. 1136) (lanes 9 to 13). Precipitates were analyzed by SDS-PAGE. For lanes 5 to 8, samples were first precipitated with anti-RelA (no. 1207) and then boiled and reprecipitated with anti-RelA (no. 1207) (lanes 5 and 7) or anti-IκBα (no. 1309) (lanes 6 and 8). Ab for IP, antibody used for immunoprecipitation. (C) 293 cells were transfected, as indicated, with plasmids encoding IκBα, full-length RelA, or full-length RelAΔNLS or were mock transfected (lane 3). Twenty-four hours later, the cells were cultured in medium containing ³⁵S-amino acids and cytoplasmic extracts were immunoprecipitated with anti-RelA (no. 1226). Precipitates were analyzed by SDS-PAGE.

An NLS-deficient mutant of human RelA (RelAΔCΔNLS) was also assayed in vivo and in vitro for its ability to bind IκBα. As expected, this mutant was predominantly cytoplasmic when overexpressed in 293 cells (Fig. 3A, panel h). Surprisingly, coexpression of IκBα with RelAΔCΔNLS such that the RelA mutant was produced in excess of IκBα resulted in strictly cytoplasmic staining of IκBα (Fig. 3A, panel i), indicating that the NLS-deficient RelA homodimer interacted with and retained IκBα in the cytoplasm. In addition to performing the in vivo experiments, we looked for an association between RelAΔCΔNLS and IκBα by immunoprecipitation. We found that anti-RelA precipitated both RelAΔCΔNLS and IκBα from lysates of cotransfected cells, indicating that the two proteins are able to associate in a stable complex (Fig. 3B, lane 8). To determine whether these unexpected results were influenced by our use of C-terminally truncated RelA, we also tested full-length RelAΔNLS. We found that it, too, could be coprecipitated with IκBα (Fig. 3C, lane 2). The same result was obtained with a human c-Rel NLS deletion mutant (c-RelΔCΔNLS); as for RelA, the c-Rel NLS is not essential for stable interaction with IκBα (Fig. 3A, panels l and m and Fig. 3B, lanes 9 to 13).

IκBα must mask both NLSs of a dimer.

We sought to determine whether a dimer needs both of its NLSs for nuclear entry or whether one is sufficient. To answer this question, we cotransfected RelAΔCΔNLS with excess p50 so that RelAΔCΔNLS would be found predominantly in heterodimers with p50. The localization of the RelA mutant was determined by immunofluorescence with anti-RelA. Despite its lack of an NLS, the RelA mutant was found in the nucleus (Fig. 3A, panel j), indicating that a single NLS is sufficient for nuclear entry of a dimer. We also observed nuclear localization of a p50 dimer in which only one monomer had an NLS (data not shown). Thus, if a dimer is to be retained in the cytoplasm, IκBα binding must mask both NLSs.

Deletion of the c-Rel or RelA NLS weakens interaction of the dimer with IκBα.

Although the NLS of c-Rel or RelA is not required for binding to IκBα, its presence does stabilize the binding. We demonstrated this several ways. First, we asked whether IκBα could inhibit the DNA-binding activity of these NLS-deficient mutants in vitro. Extracts of 293 cells expressing RelAΔCΔNLS alone or RelAΔCΔNLS together with excess IκBα were tested for DNA-binding activity by EMSA. Whereas IκBα completely inhibited DNA binding by RelAΔC (Fig. 4A, lane 2), IκBα inhibited RelAΔCΔNLS only partially (lane 4). The same results were obtained with c-RelΔC and c-RelΔCΔNLS (lanes 6 and 8). (Supershift analysis with anti-IκBα revealed no IκBα in the RelAΔCΔNLS-DNA complexes or the c-RelΔCΔNLS-DNA complexes [data not shown].) Therefore, while the NLSs of these proteins are not essential for IκBα binding, they appear to play an important role in IκBα-mediated inhibition of DNA-binding activity, suggesting that they stabilize the NFκB-IκBα complex.

FIG. 4.

Deletion of the c-Rel or RelA NLS weakens binding to IκBα. (A) 293 cells were transfected or cotransfected as indicated. Whole-cell lysates were analyzed by EMSA (top panel). Arrows point at the DNA-binding complexes. Lysates were also analyzed by immunoblotting with antisera raised against RelA (no. 1207) or c-Rel (no. 1136) (middle panel) or against IκBα (no. 1309) (lower panel). Arrows indicate detected proteins. (B) Human 293 cells were transfected with the indicated plasmid DNAs (lanes 1 to 5) or were mock transfected (lane 6). Whole-cell lysates were immunoprecipitated with anti-IκBα, and precipitates were collected on protein A-Sepharose. The washed precipitates were resuspended in 50 μl of EMSA buffer which included a ³²P-labeled oligonucleotide containing the κB site from the interleukin-6 promoter. After incubation (with agitation) at 25°C for 20 min, the samples were centrifuged at 13,000 × g for 1 min and the supernatants were analyzed by EMSA (top panel). The two lower panels are immunoblots of the whole-cell lysates probed with anti-c-Rel (upper) or anti-IκBα (lower), demonstrating that each of the plasmids was indeed expressed. Some degradation products are evident in the c-Rel panel, presumably the results of freezing and thawing of the lysate. IP, immunoprecipitate. (C) Human 293 cells were cotransfected with c-RelΔC and IκBα or c-RelΔCΔNLS and IκBα. Twenty-four hours later, the cells were cultured in medium containing ³⁵S-amino acids for 1 h. Whole-cell extracts were incubated overnight at 4°C with nonradioactive whole-cell extracts from mock-transfected cells (lanes 1 and 4) or cells transfected with c-RelΔC (lanes 2 and 5) or were incubated without extract (lanes 3 and 6). The ratio of nonradioactive to radioactive extract was about 5:1. Following incubation, the samples were immunoprecipitated with anti-IκBα, and the precipitates were analyzed by SDS-PAGE.

Second, we tested whether complexes of ΔNLS mutants with IκBα dissociate more readily than their wild-type counterparts. Using anti-IκBα, we collected complexes of c-RelΔCΔNLS and IκBα on protein A-Sepharose and incubated these complexes with a ³²P-labeled κB oligonucleotide (Fig. 4B). A significant fraction of the c-Rel mutant dissociated from the IκBα and bound to the ³²P-DNA in the supernatant (lane 5). In contrast, there was little or no detectable dissociation of wild-type c-RelΔC from IκBα (lane 4). Thus, the c-RelΔCΔNLS–IκBα complex is less stable than c-RelΔC–IκBα.

Third, we sought to determine whether excess c-RelΔC would displace c-RelΔCΔNLS from a complex with IκBα. 293 cells cotransfected with c-RelΔCΔNLS and IκBα were cultured in medium containing ³⁵S-amino acids, and lysate was incubated with a fivefold excess of nonradioactive lysate from cells transfected with c-RelΔC. After incubation for 18 h, the mixture was immunoprecipitated with anti-IκBα. A significant fraction of the ³⁵S-labeled c-RelΔCΔNLS was lost from the IκBα complex (Fig. 4C, lane 5); it had been displaced by the wild-type c-RelΔC. In contrast, ³⁵S-labeled wild-type c-RelΔC was not significantly displaced under these conditions (lane 2). Thus, as in the experiments described above, we conclude that the binding of the ΔNLS mutant to IκBα is weaker than the binding of wild-type c-RelΔC to IκBα.

Binding of IκBα to NLS mutants of p50.

To map more precisely the p50 residues required for IκBα interaction, we created a set of substitution mutants within the p50 NLS. The mutants were first tested in vivo for subcellular localization in the absence of IκBα. Their behavior is summarized in Table 1. As expected, simultaneous substitution of all four positively charged residues of the NLS (RKRQK → AAAQA) resulted in a protein that was cytoplasmic when it was expressed in 293 cells. However, the effects of individual mutations varied greatly. Mutation of the first (R362), fourth (Q365), or fifth (K366) residue to alanine had little or no effect on localization of the protein. In contrast, mutation of either K363 or R364 to alanine had a drastic effect and resulted in a cytoplasmic protein. Since K363 (unlike R364) is conserved at the same relative position in the NLSs of all Rel family proteins, we wondered if even a conservative change would be tolerated. To answer this question, we mutated K363 to arginine. Even this comparatively minor alteration gave rise to a protein that was predominantly cytoplasmic. Thus, it appears that K363 and R364 are necessary for NLS function.

TABLE 1.

NLS mutations in p50^a

Description and sequence	Localizationb
Wild type: RKRQK	N
Change all basic residues: AAAQA	C
Change in first and/or last residue
AKRQK	N
RKRQA	N, WC
AKRQA	WC, N
Change in Q: RKRAK	N
Change in K363 or R364
RARQK	C
RKAQK	C
RRRQK	C
Change in all except K363 and R364: AKRAA	C, WC

^aWild-type and mutant p50 expression plasmids were transfected into 293 cells, and the localization of p50 protein was determined by immunofluorescence. The wild-type NLS sequence is RKRQK (residues 362 to 366 of human p50). 

^bC, cytoplasmic, N, nuclear; WC, whole cell. 

We next tested the p50 NLS mutants for their abilities to interact with IκBα. p50 DNAs were cotransfected with IκBα DNA (p50 DNA in excess), and the ability of p50 protein to retain IκBα protein in the cytoplasm was determined (Fig. 5). As shown above, IκBα was distributed throughout the cell when it was transfected by itself (panel a). Cotransfection with the p50 mutant in which all of the basic NLS residues had been changed to alanine (RKRQK → AAAQA) did not alter this distribution (panel b), indicating that IκBα is not able to interact with this mutant homodimer. (Examination of the cells with anti-p50 confirmed that the p50 mutant was indeed expressed [panel c].) Thus, either deletion of the NLS or substitution of its basic residues with alanine results in the loss of p50’s ability to bind IκBα in vivo.

FIG. 5.

Binding of p50 NLS mutants to IκBα. 293 cells were transfected with IκBα alone (a) or with IκBα plus an excess of p50 NLS mutant. The cells were examined by immunofluorescence with anti-IκBα (no. 1309) or anti-p50 (no. 1141). The sequences of the mutant NLSs are shown above their respective panels. The targets of the antisera in the cotransfection experiments are underlined.

In the same way that some residues within the NLS were more important in determining localization of p50, binding to IκBα was affected by some mutations but not by others. For example, mutation of either the first or the last residue (to AKRQK or RKRQA) did not destroy the ability to bind IκBα (data not shown). Even the double mutants AKRQA and AKRQN bound IκBα and retained it in the cytoplasm (panel d). Similarly, the mutant RKRAK was able to bind IκBα (panel e), as was the triple mutant AKRAA (data not shown). These results mirror those obtained in the p50 localization experiments indicating that the mutated residues are not individually required for NLS function or for binding to IκBα.

In contrast, mutation of R364, which destroys NLS function, did not affect binding to IκBα (data not shown). Even the double mutant RKAQN was able to bind IκBα (panel f). Only K363, which was crucial for NLS function, appeared to play some specific role in binding of IκBα. With the K363 mutant, IκBα staining was predominantly cytoplasmic in some cells but distributed throughout the cell in others, indicating a lack of binding (panel g). Impaired binding was even more evident in the double mutant RARQN. Most positive cells showed IκBα throughout the cell, indicating weak or no binding by the mutant (panel h). (Staining the cells with anti-p50 verified that the mutant was expressed [panel i].) Thus, mutation of K363 to alanine adversely affects the ability to bind IκBα. In contrast to its effect on NLS function, the mutation of K363 to arginine did not impair the ability to bind IκBα (panel j), suggesting that the positive charge, rather than lysine, is what is important.

Can a homodimer with only K363 in its NLS bind IκBα? To answer this question, we generated the mutant AKAAA as well as several NLS mutants in which K363 was combined with only one other basic residue. The AKAAA mutant had little or no ability to bind IκBα (Table 2), indicating that K363 is not sufficient for IκBα binding. However, the mutants AKAAK, AKRAA, and RKAQN were able to bind IκBα, suggesting that an NLS containing K363 combined with any other basic residue is sufficient for binding.

TABLE 2.

Binding of NLS mutants of p50 to IκBα^a

Description and NLS sequence	Binds IκBα
Wild type: RKRQK	Yes
Change in K363
RARQK	Weakly
RARQN	No
RRRQK	Yes
Combination of K363 and one other basic residue
AKAAK	Yes
AKRAA	Yes
RKAQN	Yes
K363 alone: AKAAA	No

^aWild-type or mutant p50 expression plasmids were cotransfected with an IκBα expression plasmid into 293 cells (with p50 DNA in excess), and the localization of IκBα protein was determined by immunofluorescence with antiserum specific for IκBα (no. 1309). Binding to p50 was indicated by the cytoplasmic distribution of IκBα protein, and lack of binding to p50 was indicated by the whole-cell distribution of IκBα protein. 

The N-terminal region of IκBα masks the NLS of p50.

The results presented above suggest that one or more contacts between the p50 NLS and IκBα are crucial for establishing a stable protein-protein complex. A clue to the location of these contacts in IκBα is provided by the behavior of certain IκBα mutants which are missing some or all of the N-terminal domain upstream of the ankyrin repeats. For example, when we cotransfected p50 with an IκBα mutant from which residues 4 to 61 had been deleted (IκBαΔ4–61), the immunofluorescence assay revealed that the p50 was predominantly nuclear, even when the mutant IκBα was supplied in substantial excess (Fig. 6, panel b). This result was surprising, since it has been known for some time that N-terminal deletion mutants of IκBα are able to bind to p50 homodimers in vitro (13, 17). In agreement with these previous experiments, we were able to coprecipitate either p50 or c-RelΔC with IκBαΔ4–61 from lysates of cotransfected cells (Fig. 7). We also showed by immunoblotting that failure to retain p50 in the cytoplasm could not be attributed to low levels of expression of IκBα N-terminal deletion mutants. Per microgram of plasmid DNA, the expression of the IκBα mutants was comparable to that of wild-type IκBα when either was cotransfected with p50 or c-RelΔC DNA (data not shown). The immunofluorescence result suggested, therefore, that the mutant IκBα was bound to p50 in the transfected cell but that one or both p50 NLSs were unmasked, leading to nuclear localization of the p50-IκBα complex.

FIG. 6.

The IκBα N terminus masks the NLS of p50 and c-Rel. 293 cells were transfected or cotransfected, as indicated, and analyzed by immunofluorescence with anti-p50 (no. 1141), anti-c-Rel (no. 1136), or anti-IκBα (no. 1258 or no. 1309). The targets of the antisera in the cotransfection experiments are underlined. Nonunderlined partners were present in excess in all experiments. IκBΔ61 is IκBαΔ4–61, IκBΔ30 is IκBαΔ2–30, IκBΔ53 is IκBαΔ2–53, IκB (43 to 47) is IκBα with a substitution mutation in residues 43 to 47 (EQMVK → QAAAA), IκBΔ39–43 is IκBα with residues 39 to 43 deleted, IκB (48 to 52) is IκBα with a substitution mutation in residues 48 to 52 (ELQEI → QAAQA), and p50 NLS⁻ has RKAQN instead of the wild-type NLS RKRQK.

FIG. 7.

IκBαΔ4–61 coprecipitates with c-RelΔC and with p50. 293 cells were transfected as indicated. Twenty-four hours later, cells were cultured in medium containing ³⁵S-amino acids for 1 h, and cytoplasmic extracts were immunoprecipitated with the indicated antisera. Precipitates were analyzed by SDS-PAGE. The mobility of wild-type IκBα (lane 1) demonstrates that c-RelΔC and p50 coprecipitated with IκBαΔ4–61 and not with endogenous wild-type IκBα.

An alternative explanation arose from the intracellular localization of IκBαΔ4–61 in the absence of p50. In contrast to wild-type IκBα, which when transfected alone tends to be distributed throughout the cell, IκBαΔ4–61 transfected by itself was predominantly nuclear (panel c). Thus, nuclear localization of the p50-IκBαΔ4–61 complex could result either from an unmasked p50 NLS or from the intrinsic nature of the IκBα mutant. To examine these two possibilities, we used a p50 NLS mutant which retained the ability to bind IκBα but localized exclusively in the cytoplasm. If the complex between p50NLS⁻ and IκBαΔ4–61 was nuclear, we would conclude that its localization is controlled by a property of the IκBα mutant (i.e., the IκBα mutant drew the otherwise cytoplasmic p50 mutant into the nucleus). If, on the other hand, the complex was cytoplasmic, we would conclude that localization of the IκBα mutant is governed by the p50 to which it binds (i.e., the IκBα mutant is nuclear if bound to wild-type p50 but cytoplasmic if bound to a cytoplasmic NLS mutant). The second alternative proved to be accurate. Cotransfection of IκBαΔ4–61 with excess wild-type p50 resulted in nuclear localization of the IκBα mutant (panel d), while cotransfection with excess p50NLS⁻ (RKAQN) resulted in cytoplasmic localization of the IκBα mutant (panel e). Thus, nuclear localization of the complex formed between IκBαΔ4–61 and wild-type p50 must result from exposure of one or both of the p50 NLSs. That is, IκBα residues 4 to 61 either directly or indirectly shield the p50 NLS(s).

To map the critical region within the IκBα N terminus, we tested a series of deletion and substitution mutants. Deletion of IκBα residues 2 to 30 resulted in a wild-type phenotype; that is, the mutant was able to retain p50 in the cytoplasm (Fig. 6, panel f). Deletion of residues 2 to 53, however, gave the same phenotype as the deletion of residues 4 to 61, namely, the mutant was unable to retain p50 in the cytoplasm (panel g). Thus, at least a portion of the critical region must be located between IκBα residues 31 and 53. Within that region, the deletion of IκBα residues 44 to 50 resulted in the mutant phenotype (data not shown) as did the substitution of residues 43 to 47 (EQMVK → QAAAA) (panel h). The deletion of residues 39 to 43 gave the wild-type phenotype (panel i), while the substitution of residues 48 to 52 (ELQEI → QAAQA) resulted in an intermediate phenotype (partial retention of p50 in the cytoplasm [panel j]). Thus, some or all of IκBα residues 44 to 52 are involved in masking the p50 NLS(s).

When tested with IκBαΔ4–61, c-RelΔC behaved the same as p50. As with p50, even a large excess of IκBαΔ4–61 was unable to retain c-RelΔC in the cytoplasm (Fig. 6, panel k), while IκBαΔ2–30 behaved like wild-type IκBα (panel l). IκBαΔ4–61 was nuclear in the presence of excess c-RelΔC (panel m) but cytoplasmic in the presence of excess c-RelΔCΔNLS (panel n). Thus, nuclear localization of the c-RelΔC-IκBαΔ4–61 complex must result from exposed c-Rel NLS(s). As with p50, therefore, the IκBα N-terminal region either directly or indirectly masks one or both of the c-Rel NLSs.

The situation with RelA was not as clear as that with p50 and c-Rel. Mutations in the N-terminal region of IκBα definitely compromised the ability of the protein to retain RelA in the cytoplasm, but the effect was not as drastic as that with p50 or c-Rel. When cotransfected with excess IκBαΔ4–61, for example, RelA was found in the cytoplasm in some cells, throughout other cells, and in the nuclei of others (data not shown). Thus, the N terminus of IκBα interacts slightly differently with RelA than with c-Rel or p50.

DISCUSSION

In this study, we have concentrated on potential interactions between the NLSs of a Rel/NF-κB dimer and IκBα. The binding of IκBα (or the other IκBs) to a dimer somehow results in the masking of both NLSs, with the consequence that the complex is cytoplasmic. Does this masking involve direct contact between IκBα and the NLSs? Several previous studies reported that the NLS is required for IκBα binding (2, 9, 15, 25), strongly suggesting direct contact. More recently, however, it has been shown that neither the binding of IκBα to the avian v-Rel and c-Rel proteins nor the binding of Cactus to Dorsal requires the NLS (11, 23). In the present study, we show that neither human RelA nor c-Rel homodimers require NLSs to bind IκBα. The difference between earlier and more recent studies is the type of NLS mutation that was tested. Whereas most of the studies suggesting an NLS requirement were with substitution mutants, the experiments showing no requirement for the NLS were with deletion mutants. It now seems likely that the results with the substitution mutants should be interpreted not as a requirement for the NLS but, rather, as the prevention of binding due to the substituted residues. Thus, there is no direct evidence for contact between the NLSs of RelA, c-Rel or v-Rel (or Dorsal), and IκBα (or Cactus), although there is some very strong indirect evidence. First, the drastic effect of the NLS substitution mutations in blocking the binding of IκBα argues at least for their proximity. Second, deletion of the chicken c-Rel NLS rendered association with IκBα dependent on the presence of the PEST domain, suggesting that loss of the NLS had significantly weakened binding (23). Third, we showed that in the absence of an NLS in RelA or c-Rel, IκBα has a significantly reduced ability to inhibit DNA binding. Finally, we showed that complexes of IκBα with NLS deletion mutants of RelA or c-Rel were significantly less stable than complexes of IκBα with wild-type RelA or c-Rel. Therefore, it seems reasonable to conclude that IκBα contacts one or both of the RelA or c-Rel NLSs, although this contact is not required for binding.

The situation is quite different for p50, for which the NLS is required for binding to IκBα. We showed that either the deletion or the substitution of the p50 NLS resulted in the complete loss of interaction both in vitro and in vivo. The mutation of individual NLS residues revealed that K363 is required for binding, together with at least one other basic residue. Since arginine can substitute for K363, a charge-dependent interaction is suggested.

Which region of IκBα does the p50 NLS contact? We report here a new function for the N-terminal region of IκBα: it is required for masking one or both of the NLSs in a p50 or a c-Rel homodimer. The N terminus cannot be the sole IκBα contact for both NLSs because (i) the p50 NLS is required for binding to IκBα and (ii) p50 can bind IκBα in the absence of the latter’s N-terminal region. However, various alternative models are possible. For example, the IκBα N-terminal region might contact one of the NLSs of a dimer (a dispensable contact, since the complex is stable in the absence of the IκBα N terminus), while a different IκBα region masks the second NLS (a required contact, since the p50 NLS is necessary for binding). This model predicts that a p50 dimer with only one NLS would be able to bind IκBα, which is consistent with the results of our current studies. Alternatively, the IκBα N-terminal region might not contact either of the NLSs but might sterically mask one or both of them by virtue of its contact with some other region of the dimer.

In accordance with previous mutational studies (9, 11, 16, 25) and with the crystal structure of p50 and RelA (6, 6a, 10, 20), we found that IκBα interacts predominantly (or perhaps even exclusively) with the dimerization region of the RHD of p50, c-Rel, and RelA. In our experiments, this interaction was stable not only in vivo but also in vitro under conditions of moderate salt concentration and nonionic detergent. Of course, we cannot rule out the possibility that contacts are also normally made with the N-terminal region of the RHD and that these contacts might be crucial for IκBα’s ability to inhibit DNA binding by RelA- and c-Rel-containing dimers. However, several considerations should be taken into account when such hypothetical contacts are postulated. First, the ankyrin domain seems to be an unlikely source of such contacts, simply because binding is so sensitive to alteration in the ankyrin repeats (i.e., if the ankyrins cannot be altered without the loss of binding, it is unlikely that a target of the ankyrins could be altered or removed without the loss of binding). The N terminus of IκBα also now seems unlikely to contact the N-terminal region of the RHD. We demonstrate here that the N terminus of IκBα is somehow involved with the dimer NLS(s), which is suggested (but not proved) by the p50 crystal structure to be at some distance from the N-terminal domains of the monomers. Is the C-terminal PEST region of IκBα a possible source of contacts with the N-terminal region of the RHD? We know that the PEST region binds somewhere, because it is required for the inhibition of DNA binding by RelA- and c-Rel-containing dimers and because the IκBαΔPEST mutant does not bind stably to c-Rel homodimers in vitro (8). We suggested previously that this crucial contact with c-Rel might occur in the N-terminal portion of the RHD (8). If this were the case, however, deletion of the c-Rel N-terminal region would abrogate binding to IκBα in vitro just as deletion of the IκBα PEST region abolished binding to wild-type c-Rel. The results presented here, however, show that IκBα binds quite stably in vitro to c-Rel lacking its entire N-terminal region. Thus, it seems likely that the major contact with the PEST region occurs in the c-Rel dimerization region. In summary, all contacts between the dimer and IκBα may be confined to the dimerization domain.

Finally, one must be mindful of the significant difference between IκBα’s binding to p50 and its binding to c-Rel or RelA. IκBα can inhibit DNA binding by c-Rel and RelA but cannot inhibit p50. The structural basis for this functional difference is unknown but the difference suggests that discoveries regarding interaction of the IκBα with c-Rel or RelA will not always apply to interaction with p50 and vice versa.