Characterization of the Nuclear Import and Export Functions of IκBε*

Control over the nuclear localization of nuclear factor κB/Rel proteins is accomplished in large part through association with members of the inhibitor of κB (IκB) protein family. For example, the well studied IκBα protein actively shuttles between the nucleus and the cytoplasm and both inhibits nuclear import and mediates nuclear export of NF-κB/Rel proteins. In contrast, the IκBβ protein can inhibit nuclear import of NF-κB/Rel proteins but does not remove NF-κB/Rel proteins from the nucleus. To further understand how the IκB proteins control the nuclear-cytoplasmic distribution of NF-κB/Rel proteins, we have characterized the nuclear import and nuclear export functions of IκBε. Our results indicate that the IκBε protein, like the IκBα protein, actively shuttles between the nucleus and the cytoplasm. Similar to IκBα, nuclear import of IκBε is mediated by its ankyrin repeat domain and is not blocked by the dominant-negative RanQ69L protein. However, the nuclear import function of the IκBε ankyrin repeat domain is markedly less efficient than that of IκBα, with the result that nuclear shuttling of IκBε between the nucleus and the cytoplasm is significantly slower than IκBα. Nuclear export of IκBε is mediated by a short leucine-rich nuclear export sequence (NES)-like sequence (343VLLPFDDLKI352), located between amino acids 343 and 352. This NES-like sequence is required for RanGTP-dependent binding of IκBε to CRM1. Nuclear accumulation of IκBε is increased by either leptomycin B treatment or alanine substitutions within the IκBε-derived NES. A functional NES is required for both efficient cytoplasmic retention and post-induction control of c-Rel by IκBε, consistent with the notion that IκBε-mediated nuclear export contributes to control over the nucleocytoplasmic distribution of NF-κB/Rel proteins.

The NF-B/Rel transcription factor family has provided a powerful experimental paradigm for understanding how eukaryotic gene expression is regulated (reviewed in Refs. [1][2][3][4][5]. Mammalian cells have five NF-B/Rel proteins, termed p50, p52, RelA (p65), c-Rel, and RelB. The NF-B/Rel proteins are defined by the presence of a N-terminal, highly conserved region of 300 amino acids known as the Rel homology domain. The Rel homology domain is responsible for dimerization, nuclear localization, and DNA binding by NF-B/Rel proteins. Nuclear localization and DNA binding of NF-B/Rel proteins are regulated in large part through association with members of the inhibitory of B (IB) family. In mammalian cells, the major regulatory IB proteins are IB␣, IB␤, and IB⑀, which share a common domain structure. The N termini of these IB proteins constitutes a signal response domain, which is targeted for phosphorylation and ubiquitination by a variety of extracellular stimuli (6 -8). The central portion of IB proteins contains an ankyrin repeat domain comprised of six ankyrin repeats that functions as a protein-protein interaction domain (9 -13). The C-terminal regions of IB proteins are generally acidic, are important for inhibition of DNA binding by NF-B/ Rel proteins, and may regulate the basal degradation rate of IB proteins (14 -17).
Although the IB proteins were first identified as cytoplasmic inhibitors of NF-B/Rel proteins, it is now clear that IB proteins also act in the nucleus. Thus, for example, an important nuclear function for IB␣ is post-induction repression of NF-B activation (18 -32). Increased nuclear localization of IB␣ has been demonstrated after cytokine-induced degradation of the cytoplasmic pool of IB␣ proteins (22,25). Nuclear import of IB␣ is mediated by sequences within the ankyrin repeat domain and does not require the importin-␣/importin-␤ nuclear import complex utilized by conventional nuclear localization sequences (NLSs) 1 (26). In the nucleus, IB␣ is able to displace NF-B/Rel proteins from their cognate DNA elements and direct the export of NF-B from the nucleus to the cytoplasm (18 -29). IB␣ binds to the nuclear export receptor CRM1 via a conventional nuclear export sequence (NES) within its N-terminal domain and functions as an adaptor protein to bridge CRM1 and NF-B/Rel proteins (20,28). Although the importance of IB␣-mediated nuclear export of NF-B/Rel proteins in controlling NF-B/Rel-dependent transcription is not fully understood, one consequence of IB␣mediated nuclear export of NF-B/Rel proteins is to regenerate the cytoplasmic pool of NF-B in preparation for subsequent reactivation (30).
The IB␤ protein has also been detected in the nucleus (10,27,31). In contrast to IB␣, IB␤ forms a ternary complex with NF-B and DNA and is thought to protect NF-B from inactivation by IB␣ in the nucleus (10,27). A further contrast between IB␣ and IB␤ is the absence of a functional NES in IB␤ (20). IB␣-mediated cytoplasmic retention of NF-B/Rel proteins is disrupted by the CRM1-specific inhibitor, leptomycin B, whereas IB␤-mediated cytoplasmic retention of NF-B/ Rel proteins is independent of CRM1 (16,32). Taken together, these data suggest that IB␣ and IB␤ have distinct and nonoverlapping nuclear functions.
The third major IB protein, IB⑀, is generally regarded as a cytoplasmic protein. However, the possibility that IB⑀ can shuttle between the cytoplasm and the nucleus has not been examined. We now report that the IB⑀ protein is able to shuttle between the cytoplasm and the nucleus. Nuclear import of IB⑀ is mediated by its ankyrin repeat domain, although the IB⑀-derived ankyrin repeat domain is less efficient at promoting nuclear localization of a heterologous reporter protein than the ankyrin repeat domain of IB␣. Similar to IB␣, nuclear import of IB⑀ is not blocked by a dominant-negative Ran protein that otherwise disrupts classical NLS-dependent nuclear import. The IB⑀ protein interacts with CRM1 in a RanGTPdependent manner and contains a canonical "leucine-rich" NES-like motif (amino acids 343-352) that is required for Ran-GTP-dependent binding to CRM1. Mutation of this NES-like motif in IB⑀ results in increased nuclear localization of IB⑀. Importantly, the C-terminal NES-like motif of IB⑀ is required for both efficient cytoplasmic retention and post-induction repression of c-Rel by IB⑀. Taken together, our data indicate that IB⑀ is not restricted to the cytoplasm as previously suggested but actively shuttles between the nucleus and the cytoplasm. Furthermore, our results suggest that one of the functions of IB⑀ in the nucleus is to direct the export of NF-B/Rel proteins back to the cytoplasm.

EXPERIMENTAL PROCEDURES
Construction of Recombinant DNA Molecules-The construction of DNA molecules was performed by standard techniques (33). The cytomegalovirus (CMV)-derived vector encoding human IB⑀ was obtained from Nancy Rice (11). Point mutations of IB⑀ were constructed from phagemid single-strand DNA and sequenced to confirm the presence of desired mutation.
Deletion mutants of IB⑀ and human c-Rel were generated using PCR with appropriate oligonucleotide primers, and the PCR-derived DNAs were cloned into CMV-based mammalian expression vectors as HindIII-XhoI fragments. All GST expression vectors for IB⑀ were constructed using PCR with appropriate oligonucleotide primers, and the PCR-derived DNAs were cloned into pGEX-5X-2 vector as EcoRI-XhoI fragments. The His-tagged IB protein expression vectors were constructed using PCR with appropriate primers, and the PCR-derived DNAs were inserted into the pET15b vector as NdeI-XhoI fragments. The Myc-IB⑀ protein expression vectors were constructed using PCR, and PCR-derived DNAs were cloned into a CMV-derived 5Ј Myc epitopetagged vector as BamH1 and XhoI. The Myc-NPc-ARD fusion protein containing the IB␣ ARD (amino acids 59 -243) has been described previously (29). The Myc-NPc-ARD fusion protein containing the IB␤ (amino acids 47-309) and IB⑀ ARD (amino acids 109 -339) was constructed using PCR with appropriate primers, and the PCR-derived DNA was cloned into a CMV-derived pMyc-NPc expression vector (29). Expression vectors for GST-NLS, RanQ69L, RanBP1, and HA-CRM1 have been described previously (26,28). Bacterial expression vectors for importin-␣ and importin-␤ were obtained from Dirk Gorlich (University of Heidelberg).
Expression and Purification of Recombinant Proteins-The recombinant GST-IB⑀ proteins were expressed in Escherichia coli strain BL21(DE3)pLysS and purified with glutathione-agarose (26).
The RanQ69L, RanBP1, importin-␤, importin-␣, and CRM1 proteins were expressed as His-tagged proteins in E. coli strain BLR pRep4 and purified using metal ion affinity chromatography (26). Other Histagged proteins were expressed in BL21(DE3)pLysS cells and purified by metal-chelate affinity chromatography. HA-CRM1 was purified from insect cells infected with a baculovirus vector (28). The labeling of proteins with fluorescein was carried out as described previously (26).
Cell Culture and Transfection-Either COS1 cells or 3T3 cells derived from IB␣ nullizygous mice were grown in Dulbecco's modified eagle medium containing 10% fetal bovine serum. Transfections were performed with 1 g of the indicated plasmids and 15 l of Lipo-fectAMINE (Invitrogen)/35-mm dish. The cellular localization of c-Rel, IB⑀, Myc-tagged IB⑀, and the Myc-NPc-ARD fusion proteins were analyzed 40 h after transfection by indirect immunofluorescence as previously described (28). The cells were fixed with ice-cold methanol and washed with phosphate-buffered saline before antibody incubations. For TNF-␣ treatment, the transfected 3T3 cells derived from IB␣ nullizygous mice were refed with complete medium (Dulbecco's modified eagle medium containing 10% fetal bovine serum) containing cycloheximide (100 g/ml; Sigma) and TNF-␣ (10 ng/ml; Chemicon International) for 2 h. The transfected cells were washed three times with Dulbecco's modified Eagle's medium and incubated with complete medium either lacking or containing 20 nM leptomycin B for 3 h before fixation and analysis by double-label indirect immunofluorescence. For detection of the IB⑀ proteins, a rabbit polyclonal anti-IB⑀ antibody (Santa Cruz Biotechnology) was followed with a fluorescein isothiocyanate-conjugated anti-rabbit antibody (Jackson Laboratories). For detection of either c-Rel or the Myc-NPc-ARD fusion proteins, mouse monoclonal antibodies directed against either c-Rel (Santa Cruz) or the Myc epitope (Santa Cruz) were followed by a Cy3-conjugated anti-mouse antibody (Jackson Laboratories). For detection of c-Rel and Myc-tagged IB⑀ in 3T3 cells derived from IB␣ nullizygous mice, a rabbit polyclonal antibody directed against c-Rel (Santa Cruz) and a mouse monoclonal antibody directly against the Myc epitope (Santa Cruz) were followed by a fluorescein isothiocyanate-conjugated anti-rabbit and a Cy3-conjugated anti-mouse antibody, respectively.
In Vitro Binding Assay-To measure binding of GST-IB⑀ to immobilized CRM1, anti-HA-agarose beads containing 100 ng of HA-CRM1 was incubated with 2 g of GST-IB⑀ and 2 g of RanQ69L GTP in binding buffer (10 mM sodium phosphate, pH 7.3, 150 mM NaCl, 0.1% Triton X-100, and 1 mM 2-mercaptoethanol) and rotated at 4°C for 2 h. The matrix was extensively washed with binding buffer before the addition of SDS-PAGE sample buffer. To measure binding of HA-CRM1 to immobilized IB⑀ proteins, 2 g of each GST-IB⑀ protein was loaded onto glutathione-agarose beads in binding buffer at 4°C for 30 min. The beads were extensively washed and incubated with 20 l of insect cell lysates containing HA-CRM1 in the presence of 2 g of RanQ69L-GTP at 4°C for 2 h. The matrix was extensively washed with binding buffer before the addition of SDS-PAGE sample buffer. In all binding assays, immunoblot analysis was used to measure proteins that remained bound to the respective solid supports after extensive washing. The samples were electrophoresed through SDS-polyacrylamide gels and transferred to nitrocellulose. HA-CRM1 was detected with a mouse monoclonal antibody against the HA epitope (Covance), GST-IB⑀ proteins were detected with a mouse monoclonal antibody against GST (Santa Cruz), and the RanQ69L protein was detected with a rabbit polyclonal antibody against Ran (Covance). All primary antibodies were used at dilutions of 1:2000 -1:4000. For all immunoblot analyses, the secondary antibodies were either horseradish peroxidase-coupled antimouse IgG (New England Biolabs) used at a dilution of 1:20,000 or horseradish peroxidase-coupled anti-rabbit IgG (Sigma) used at a dilution of 1:10,000. To allow detection of multiple proteins on the same nitrocellulose filter, some filters were stripped with IgG elution buffer (Pierce) before subsequent antibody incubations. The proteins were visualized using Amersham Biosciences ECL reagents followed by exposure to XAR-5 film. RanGAP Protection Assay-RanGAP protection assay was performed as described previously except when using His-tagged IB proteins purified from E. coli (28). In brief, 0 -15 M His-tagged IB proteins were incubated in the presence of 100 pM Ran[␥-32 P]GTP and 1 M CRM1 at 15°C for 30 min. The samples were placed on ice, 10 nM RanGAP was added, and the samples were immediately placed at 30°C for 2 min. The amount of [␥-32 P] GTP bound to Ran was measured by the filter binding assay (28).
Immunoprecipitation Analysis of IB⑀ and c-Rel Proteins-Cell lysates for the immunoprecipitation experiments were prepared in ELB (50 mM Tris-HCl, pH 7.9, 250 mM NaCl, 0.1% Triton X-100, 5 mM EDTA, and 1 mM dithiothreitol) containing 1 mM phenylmethylsulfonyl fluoride and a protease inhibitor mixture (Calbiochem). To inhibit phosphatase activity, 0.4 mM sodium orthovanadate and 1 mM sodium fluoride were included in the lysis buffer. Expression of c-Rel and of the Myctagged IB⑀ proteins was confirmed by immunoblot analysis using a rabbit polyclonal antibody directed against c-Rel and or a mouse monoclonal antibody directed against the Myc epitope (Santa Cruz). Equivalent amounts of protein lysates were used for the immunoprecipitation analysis. Immunoprecipitation of Myc-tagged IB⑀ proteins was performed with 20 l of anti-Myc-agarose beads/sample (Santa Cruz). The presence of c-Rel in the immunoprecipitates was determined by immunoblot analysis.
In Vitro Nuclear Import Assay-HeLa cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. Approximately 16 h before the transport assays, 6 ϫ 10 5 HeLa cells were placed onto 35-mm diameter plates containing glass coverslips. Cells on coverslips were permeabilized with 50 g of digitonin/ml (Calbiochem) in transport buffer for 5 min on ice. The transport reactions were conducted for 20 min at room temperature. A standard 50-l transport reaction contained an energy-regenerating system (1 mM ATP, 0.1 mM GTP, 5 mM creatine phosphate, 20 units of creatine phosphokinase/ml), protease inhibitor mix, 2 mM dithiothreitol, and 15 l of rabbit reticulocyte lysate. The fluorescein-labeled import substrates were added to a final concentration of 100 g/ml. After nuclear transport reactions, the coverslips were fixed with 4% paraformaldehyde and mounted onto glass slides with Mowiol containing 2.5% DABCO (1,4diazabicyclo[2.2.2]octane; Sigma). When indicated, HeLa cells growing on coverslips were treated with 10 nM leptomycin B for 30 min before digitonin permeabilization, and 10 nM leptomycin B was included throughout the transport reactions. The CRM1 and RanBP1 dependence experiments were performed in the presence of 2 g each of CRM1 and RanBP1. The RanQ69L experiments were performed in the presence of 5 g of RanQ69L preloaded with GTP. The GST-NLS nuclear import reactions were performed in the presence of 3 g each of importin-␣ and importin-␤. The panels shown in Fig. 2 were taken with a 60ϫ oil immersion lens on a Nikon Optiphot-2 equipped with a Diagnostic Imaging Spot II camera. Images were collected using software from Diagnostic Imaging and transferred to Adobe Photoshop. No manipulation of the images was performed except for converting the original color images to black and white. Equivalent exposure time periods were used for all panels in Fig. 2.

Nuclear Import of IB⑀ in Digitonin-permeabilized HeLa
Cells-The IB⑀ protein is located in the cytoplasm, yet shares along with IB␣ and IB␤ a central ankyrin repeat domain that is responsible for nuclear import of IB␣ ( Fig. 1) (29). To determine whether IB⑀ is capable of entering the nucleus, in vitro nuclear import assays were performed using fluoresceinlabeled IB⑀ and digitonin-permeabilized HeLa cells. In the presence of reticulocyte lysate and an energy regenerating system, nuclear accumulation of IB⑀ was readily observed ( Fig. 2A). Nuclear accumulation of IB⑀ was energy-dependent and was enhanced in the presence of reticulocyte lysate (data not shown).
Nuclear Import of IB⑀ Is Independent of GTP Hydrolysis by Ran-The directionality of nuclear export and nuclear import pathways is determined by the asymmetric distribution of the GTP-bound and GDP-bound forms of Ran between the nucleus and the cytoplasm. To determine whether nuclear import of IB⑀ is sensitive to perturbation of the asymmetric Ran-nucleotide gradient, nuclear accumulation of IB⑀ was determined in the presence of the dominant-negative RanQ69L protein bound to GTP. RanQ69L GTP efficiently blocked nuclear import of a protein containing a classical NLS (GST-NLS; Fig. 2J). In contrast, RanQ69L GTP did not block nuclear import of IB⑀ (Figs. 2, D and H). Thus, like IB␣ and in contrast to nuclear import of proteins that contain a classical NLS, nuclear import of IB⑀ is not disrupted by perturbation of the asymmetric Ran-GTP gradient.
Because nuclear export reactions can also be monitored using the digitonin-permeabilized cell assay, the possibility that IB⑀ may also undergo nuclear export was examined. Nuclear accumulation of IB⑀ was abolished when the nuclear import reactions were carried out in the presence of CRM1 and RanBP1 (Fig. 2B). In contrast, the inclusion of leptomycin B in the import reaction enhanced nuclear accumulation of IB⑀, even in the presence of exogenous CRM1 (Fig. 2C).
The ability of exogenous CRM1 to block nuclear accumulation of IB⑀ in digitonin-permeabilized HeLa cells suggested that IB⑀ can be exported from the nucleus via the CRM1-dependent nuclear export pathway. To define the region(s) of IB⑀ required for CRM1-dependent nuclear export, several mutant IB⑀ proteins were characterized for their ability to accumulate in the nucleus of digitonin-permeabilized cells in a CRM1independent manner (data not shown). Removal of 21 amino acids from the C terminus of IB⑀ was sufficient to allow CRM1-independent nuclear accumulation of IB⑀ (His-IB⑀ 1-340 ; Figs. 2, E and F).
IB⑀ Binds to CRM1 in a RanGTP-dependent Manner-The formation of an export-competent complex between CRM1 and a NES-bearing cargo protein requires cooperative binding of RanGTP (34). To determine whether IB⑀ binds to CRM1 in a RanGTP-dependent manner, the ability of purified IB⑀ protein to bind CRM1 was determined in the absence or presence of RanGTP. For this purpose, GST-tagged IB⑀ (GST-IB⑀-wt) was prepared. HA-tagged CRM1 (HA-CRM1) was immobilized on anti-HA-agarose beads, and GST-IB⑀ was added in the presence or absence of RanQ69L-GTP. To closely mimic the in vivo situation where CRM1 interacts with multiple cargo proteins, lysates from 293 cells were included during binding reaction. After incubation, the beads were extensively washed, and bound proteins were determined with immunoblot analysis. Under these conditions, binding of IB⑀ to CRM1 was substantially increased in the presence of RanQ69L GTP (Fig.  3A, compare lane 1 and 2).
IB⑀ Contains a Canonical NES-like Motif in Its C Terminus That Is Required for RanGTP-dependent Binding to CRM1-Because removal of 21 amino acids from the C terminus of IB⑀ was sufficient to allow CRM1-independent nuclear accumulation of IB⑀ (His-IB⑀ 1-340 ; Figs. 2, E and F), the possibility of a NES in this region of IB⑀ was carefully investigated. Indeed, the amino acids 343-352 of IB⑀ contain a putative leucine-rich NES-like motif, including the hallmark "core" tetramer found in other classical NESs (Fig. 1). To determine whether this putative NES-like motif in IB⑀ is required for binding to CRM1, two mutant GST-IB⑀ proteins were prepared, a C-terminal deletion mutant (GST-IB⑀ 1-340 ) and a point mutant containing alanine substitutions for leucine 350 and isoleucine 352 (GST-IB⑀ L350A/I352A ) of IB⑀. Binding of these mutant IB⑀ proteins to CRM1 in the presence of RanGTP was determined. Binding of both mutant GST-IB⑀ proteins to CRM1 in the presence of RanQ69L GTP was significantly reduced compared with wild-type GST-IB⑀ (Fig. 3A,  compare lanes 2-4). To confirm this observation, the respective GST-IB⑀ proteins were immobilized on glutathione-agarose and incubated with insect cell lysates containing HA-CRM1. An equivalent amount of RanQ69L GTP was added to each sample. The agarose beads were extensively washed, and bound HA-CRM1 was determined with immunoblot analysis. Binding of HA-CRM1 to the immobilized GST-IB⑀ L350A/Il352 and GST-IB⑀ 1-340 proteins was markedly reduced compared with wild-type GST-IB⑀ (Fig. 3B, compare lane 3-5).
Quantitative Analysis of IB-CRM1 Interactions-A Ran-GAP protection assay was used to provide a quantitative measure of the affinity of IB for CRM1. The slow intrinsic GTPase activity of Ran is markedly stimulated by the GTPase-activating protein, RanGAP. However, upon formation of an exportcompetent ternary complex between RanGTP, CRM1, and a NES-bearing cargo protein, RanGAP is no longer able to stimulate the GTPase activity of Ran. In this assay, the ability of a NES-bearing cargo protein to inhibit RanGAP-stimulated GTP hydrolysis in a concentration-dependent manner provides an estimate of the affinity of the cargo protein for CRM1.
His-tagged IB␣, IB␤, and IB⑀ proteins were assayed for the inhibition of RanGAP-stimulated GTP hydrolysis by Ran. 8 M His-IB␣ provided 50% protection against RanGAP-stimulated GTP hydrolysis. His-IB⑀ protein was equally effective as His-IB␣ protein for protection against RanGAP. In contrast, 15 M His-IB␤ protein did not provide significant protection against RanGAP-stimulated GTP hydrolysis by Ran (Fig. 4A).
The ability of the mutant His-IB⑀ proteins to inhibit Ran-GAP-stimulated GTP hydrolysis by Ran was determined. 15 M His-IB⑀ 1-340 protein did not provide significant protection against RanGAP-stimulated GTP hydrolysis by Ran, although 8 M His-IB⑀ L350A/I352A protein provided 20% protection, suggesting that His-IB⑀ L350A/I352A protein has an ability of partial interaction to CRM1 (Fig. 4B).
Taken together, the binding assays and the RanGAP protection assays provide strong evidence that the C-terminal NESlike sequence of IB⑀ is required for formation of an exportcompetent complex with CRM1 and RanGTP. Furthermore, IB␣ and IB⑀ proteins have comparable affinities to CRM1. In contrast, under this experimental condition, IB␤ does not have a measurable affinity to CRM1.
The C-terminal NES-like Motif Is Required for Cytoplasmic Localization of IB⑀-To determine whether the NES motif in IB⑀ contributes to the cytoplasmic localization of IB⑀, expression vectors encoding either wild-type or mutant IB⑀ proteins were transiently transfected into COS1 cells. Localization of the ectopically expressed IB⑀ proteins was determined by indirect immunofluorescence. The wild-type IB⑀ protein was predominantly localized to the cytoplasm ( Fig. 5A; Table I). In contrast, the mutant IB⑀ proteins displayed a whole-cell pattern of staining (Figs. 5, C and D). Furthermore, the staining pattern of the wild-type IB⑀ protein shifted from a cytoplasmic pattern to a whole-cell pattern after a 3-h treatment with leptomycin B (Fig. 5B). The staining pattern of the IB⑀ mutant proteins was not altered by leptomycin B treatment (Table  I). Taken together, our results indicate that the C-terminal NES-like motif of IB⑀ contributes to cytoplasmic localization of IB⑀.
Inefficient Nuclear Import of IB⑀-The relatively slow nuclear accumulation of IB⑀ in leptomycin B-treated cells is in contrast to the rapid nuclear accumulation of IB␣ after leptomycin B treatment (21,25). To determine whether this difference between IB␣ and IB⑀ is a consequence of impaired nuclear import of IB⑀ relative to IB␣ or reflects the ability of IB⑀ to access a CRM1-independent nuclear export pathway,   we compared the nuclear import function of the respective ARDs from IB␣ and IB⑀ along with the ARD from IB␤. The ankyrin repeat domains of each of the IB proteins were fused to the core domain of nucleoplasmin containing an N-terminal Myc epitope tag (Myc-NPc). Expression vectors encoding the respective Myc-NPc-ARD fusion proteins were transfected into COS1 cells, and the ability of the respective fusion proteins to localize to the nucleus was determined. As previously reported, the IB␣ ARD efficiently relocalized Myc-NPc from the cytoplasm to the nucleus (Figs. 6, A and B; Table II) (29). The IB⑀ ARD also redistributed Myc-NPc from the cytoplasm to the nucleus, although with less efficiency compared with the ARD of IB␣ (Fig. 6C; Table II). In contrast, the ARD of IB␤ did not significantly relocalize Myc-NPc from the cytoplasm to the nucleus ( Fig. 6D; Table II). Thus, the relatively slow nuclear accumulation of IB⑀ in leptomycin B-treated cells is likely due to the relatively weak nuclear import function of the IB⑀ ARD.
The C-terminal NES Motif of IB⑀ Is Required for Cytoplasmic Localization of NF-B/Rel Proteins-To determine the effect of IB⑀ on localization of NF-B/Rel proteins, a CMVderived expression vector encoding c-Rel was cotransfected with expression vectors encoding wild-type or mutant IB⑀ proteins into COS1 cells. For this purpose, a C-terminal truncated c-Rel comprising amino acids 1-362 was utilized. The c-Rel 1-362 -truncated c-Rel proteins lack a C-terminal transcriptional activation domain. Importantly, the c-Rel 1-362 protein still associates with both wild-type and mutant IB⑀ proteins (Fig. 7).
This C-terminal truncated c-Rel protein was strongly localized to the nucleus when ectopically expressed in COS1 cells (Table I). As expected, coexpression of c-Rel 1-362 with the wildtype IB⑀ proteins resulted in the colocalization of c-Rel  and IB⑀ in the cytoplasm (Figs. 8, A and B). In 15% of cells treated with 20 nM leptomycin B, both c-Rel 1-362 and IB⑀ marginally translocated toward the nucleus, although a significant portions of c-Rel 1-362 and IB⑀ remained in the cytoplasm (Figs. 8, C and D, Table I). To determine the extent to which the IB⑀ NES contributes to cytoplasmic retention of c-Rel by IB⑀, expression vectors for either IB⑀ 1-340 or IB⑀ L350A/Il352 were cotransfected along with an expression vector for c-Rel 1-362 into COS1 cells, and the localization of the c-Rel 1-362 protein was monitored by indirect immunofluorescence. Nuclear staining of the c-Rel 1-362 protein was readily detected despite coexpression of the mutant IB⑀ proteins (Figs. 8, E, F, I, and J; Table I Table I).
IB␣ is one of the major determinants of nucleocytoplasmic localization of NF-B/Rel proteins. Therefore, the endogenous IB␣ protein might affect the nucleocytoplasmic localization of ectopically expressed c-Rel 1-362 in COS1 cells considering a marginal nuclear translocation of c-Rel 1-362 and IB⑀ proteins was observed in leptomycin B-treated COS1 cells (Table I).
To avoid this possibility, 3T3 cells derived from IB␣ nullizygous mice were used for cotransfection assay. Coexpression of c-Rel 1-362 with the Myc-tagged wild-type IB⑀ proteins resulted in the colocalization of c-Rel 1-362 and Myc-IB⑀ in the cytoplasm (Figs. 9, A and B). In contrast, nuclear staining of the c-Rel 1-362 protein was readily detected despite coexpression of the Myc-tagged mutant IB⑀ proteins (Figs. 9, C-F, Table  III).
To examine the possibility that IB⑀ nuclear export might contribute to post-induction repression of c-Rel, 3T3 cells de-  Table III).
Under the same condition, the nucleocytoplasmic distribution of the c-Rel 1-362 protein and of the Myc-tagged mutant IB⑀ proteins was insensitive to leptomycin B (Figs. 10, E-L, Table  III). DISCUSSION The IB⑀ protein, first identified in a two-hybrid screen using the RelA subunit of NF-B as bait, is emerging as an important regulator of NF-B/Rel proteins (11,35,36). Most notably, loss of IB⑀ results in specific defects in T-cell development, Ig isotype switching, and cytokine gene expression (37). IB⑀ has been suggested to act exclusively in the cytoplasm to inhibit nuclear localization of NF-B/Rel proteins (36), although nucleocytoplasmic shuttling of IB⑀ has not been examined in detail. In this report, we provide evidence that IB⑀ contains nuclear import and export functions that enable shuttling of IB⑀ between the nucleus and the cytoplasm. Furthermore, IB⑀ is able to mediate CRM1-dependent nuclear export of NF-B/Rel proteins.
Our results provide several lines of evidence for a nuclear import function in IB⑀. First, although IB⑀ displays a predominant cytoplasmic pattern of localization, disruption of nuclear export with the CRM1-specific inhibitor, leptomycin B, results in a significant relocalization of IB⑀ to the nucleus. Second, fusion of the ankyrin repeat domain of IB⑀ is able to target a heterologous protein to the nucleus. Finally, IB⑀ is able to accumulate in the nucleus of digitonin-permeabilized cells in standard in vitro nuclear import assays. Taken together, our results indicate that the ankyrin repeat domain of IB⑀, like the ankyrin repeat domain of IB␣, is a functional nuclear import domain. It is likely that IB⑀ and IB␣ utilize a similar mechanism for transit through the nuclear pore complex, as nuclear import of IB⑀, like that of IB␣, is not disrupted by the dominant-negative RanQ69L-GTP protein (26).
Despite the overall similarities in nuclear import between IB␣ and IB⑀, one important difference is the relative effectiveness of the ankyrin repeat domains of IB␣ and IB⑀ at promoting nuclear import. Fusion of the ankyrin repeat domain of IB␣ to the core domain of nucleoplasmin results in a marked relocalization of nucleoplasmin from the cytoplasm to the nucleus. The ankyrin repeat domain of IB⑀ is also effective at relocalization of the nucleoplasmin core from the cytoplasm to the nucleus and the cytoplasm, although it is less efficient than the ARD of IB␣. In contrast, the ARD of IB␤ is ineffective at relocalization of nucleoplasmin core to the nucleus. This finding suggests that although the ankyrin repeat domains of IB proteins share overall structural similarities, the ankyrin repeat domains of each IB proteins are markedly different in their abilities of mediating nuclear import.
The relative weakness of the IB⑀ ankyrin repeat as a functional nuclear import signal is also reflected in the behavior of the full-length IB⑀ protein in leptomycin B-treated cells. In the case of IB␣, 30 min of leptomycin B treatment is sufficient to relocalize the bulk of IB␣ to the nucleus (21,25), whereas only a partial relocalization of IB⑀ is achieved even after several hours of leptomycin B. Several other reports also demonstrate that nuclear localization of IB⑀ is markedly dimin-ished relative to IB␣, leading to the prevailing suggestion that IB⑀ functions only in the cytoplasm (20,36). Our results agree with the notion that nuclear import of IB⑀ is less efficient than IB␣. Nevertheless, the ability of IB⑀ to shuttle between the cytoplasm and the nucleus suggests that IB⑀ may function in both compartments to regulate NF-B/Rel proteins.
One potential nuclear function of IB⑀ might be to mediate nuclear export of NF-B/Rel proteins. In agreement with the results of Tam and Sen (38), we find that inhibition of CRM1dependent nuclear export only marginally alters the steadystate cytoplasmic localization of the c-Rel/IB⑀ complex. However, we find that CRM1-dependent nuclear export is required for efficient IB⑀-mediated cytoplasmic retention of c-Rel after TNF-␣/cycloheximide-induced degradation of IB⑀. We have identified a canonical NES in IB⑀ that is required for RanGTPdependent binding of IB⑀ to CRM1. Furthermore, IB␣ and IB⑀ proteins have comparable affinities to CRM1. In contrast, IB␤ does not have measurable affinity to CRM1. The NES of IB⑀ is also required for cytoplasmic localization of both IB⑀ and for IB⑀-dependent cytoplasmic retention of c-Rel. These results suggest that removal of NF-B/Rel proteins from the nucleus is an important nuclear function of IB⑀.
Post-induction repression of activated gene expression provides an effective mechanism for temporal regulation of gene expression. In the NF-B/Rel system, it is clear that a major function of the IB␣ protein is to participate in an autoregulatory circuit that attenuates NF-B-dependent transcription. Thus, for example, the IB␣ gene is rapidly induced by NF-B/ Rel proteins, and IB␣ is able to efficiently enter the nucleus, displace NF-B/Rel proteins from DNA, and mediate their removal from the nucleus (18 -29). Our present results together with the previous demonstration that expression of the IB⑀ gene is induced by NF-B/Rel proteins (11) support the notion that IB⑀ also participates in post-induction repression of NF-B/Rel proteins. However, as suggested by Whiteside et al. (11), it is likely that that IB␣ and IB⑀ have distinct roles in the post-induction repression of NF-B/Rel proteins. IB⑀ expression is induced with markedly slower kinetics than IB␣, and nuclear import of IB⑀ is less efficient than IB␣ (11,35,36). IB␣ and IB⑀ both share the ability to inhibit DNA binding of NF-B/Rel proteins and to direct their removal from the nucleus, although these IB proteins have distinct specificities for different NF-B/Rel proteins (11). A plausible scenario is that IB␣ controls expression of NF-B-dependent genes that must be activated and subsequently repressed in a relatively short time frame, whereas IB⑀ is specialized for control over NF-B-dependent genes that are activated and subsequently repressed over a longer time period.