IκBα and IκBα/NF-κB Complexes Are Retained in the Cytoplasm through Interaction with a Novel Partner, RasGAP SH3-binding Protein 2*

IκBα inhibits the transcriptional activity of NF-κB both in the cytoplasm by preventing the nuclear translocation of NF-κB and in the nucleus where it dissociates NF-κB from DNA and transports it back to the cytoplasm. Cytoplasmic localization of inactive NF-κB/IκBα complexes is controlled by mutual masking of nuclear import sequences of NF-κB p65 and IκBα and active CRM1-mediated nuclear export. Here, we describe an additional mechanism accounting for the cytoplasmic anchoring of IκBα or NF-κB/IκBα complexes. The N-terminal domain of IκBα contains a sequence responsible for the cytoplasmic retention of IκBα that is specifically recognized by G3BP2, a cytoplasmic protein that interacts with both IκBα and IκBα/NF-κB complexes. G3BP2 is composed of an N-terminal domain homologous to the NTF2 protein, followed by an acidic domain sufficient for the interaction with the IκBα cytoplasmic retention sequence, a region containing five PXXP motifs and a C-terminal domain containing RNA-binding motifs. Overexpression of G3BP2 directly promotes retention of IκBα in the cytoplasm, indicating that subcellular distribution of IκBα and NF-κB/IκBα complexes likely results from a equilibrium between nuclear import, nuclear export, and cytoplasmic retention. The molecular organization of G3BP2 suggests that this putative scaffold protein might connect the NF-κB signal transduction cascade with cellular functions such as nuclear transport or RNA metabolism.

Rel/NF-B transcription factors play a major role in inducible expression of a number of cellular genes involved in immune, inflammatory, and anti-apoptotic responses (1-3). Human NF-B is composed of a homo-or heterodimer of proteins that belong to the multigene family of transcription factors comprising p50, p52, p65/RelA, c-Rel, and RelB (4 -12). The prototypical NF-B is a heterodimeric p50/p65 molecule. Each member of NF-B/Rel family of proteins contains a Rel homology domain that is responsible for nuclear translocation, dimerization, and sequence-specific DNA binding. In most unstimulated cells, NF-B is retained in an inactive form in the cytoplasm through its association with the IB inhibitor pro-teins (13)(14)(15)(16). IBs also belong to a multigene family of proteins including IB␣, IB␤, IB⑀, Bcl-3, and also the C-terminal domains of p50 and p52 precursors (p105 and p100, respectively) that in isolation are known as IB␥ and IB␦, respectively (17)(18)(19)(20)(21)(22)(23)(24)(25). Members of the IB family contain multiple conserved ankyrin repeat domains that interact with NF-B factors such that their nuclear localization sequences (NLS) 1 are masked, leading to cytoplasmic retention of the complex. IB proteins are also characterized by their ability to inhibit NF-B DNA binding activity.
IB␣ is composed of a surface-exposed N-terminal domain, a central region containing six ankyrin repeat domains, and a highly acidic C-terminal domain. Upon stimulation of cells with appropriate signals such as tumor necrosis factor or interleukin 1, a signaling cascade is initiated leading to activation of two IB␣ kinases, IKK-1 and IKK-2, which phosphorylate IB␣ on Ser-32 and Ser-36 (26 -30). After phosphorylation, IB␣ is polyubiquitinated on Lys-21 and Lys-22 and degraded by the 26 S proteasome (31)(32)(33). NF-B can then translocate to the nucleus where it activates transcription of responsive genes including that of its inhibitor, IB␣ (34 -37). Newly synthesized IB␣ accumulates in the cytoplasm but also in the nucleus where it dissociates NF-B from DNA and transports NF-B back to the cytoplasm (38,39). This latter function is ensured by leucine-rich nuclear export sequences (NES) located in the C-terminal (amino acids 265-277; Refs. 39 -41) and N-terminal (amino acids [45][46][47][48][49][50][51][52][53][54] domains of IB␣. NES are recognized by the CRM1 receptor (exportin 1) that promotes the nuclear export of NES-containing proteins in conjunction with Ran-GTP (46 -51). In some cells such as HeLa cells or peripheral blood T lymphocytes, IB␣ is expressed not only in the cytoplasm but also in the nucleus in the absence of stimulation (52,53). Nuclear IB␣ is not sensitive to signal-induced degradation. An efficient nuclear export of IB␣ is therefore essential not only in the post-induction repression of transcription but also for maintaining a low level of IB␣ in the nuclear compartment, thus allowing NF-B to be transcriptionally active upon cell activation (43,53).
The inhibitory function of IB␣ in the nucleus is primarily limited by its ability to translocate into this compartment. In this respect, it has been clearly shown that ankyrin repeats of IB␣, which interact with NF-B are also responsible for the nuclear import of IB␣. More precisely, a region accounting for the nuclear translocation of IB␣ has been identified in the second ankyrin repeat (54,55). Co-crystallization of NF-B/ IB␣ complexes indicates that nuclear import sequences of p65 and IB␣ proteins are mutually masked, whereas the NLS of p50 is exposed (56,57), indicating that an additional mechanism accounts for the cytoplasmic localization of the complex NF-B/IB␣. It has been recently reported that an NES located in the N-terminal domain of IB␣ (which was not in the NF-B/IB␣ structure) participates to the cytoplasmic localization of inactive NF-B/IB␣ complexes (43,44) by inducing an efficient export of the complex out of the nucleus. In addition, the N-terminal region of IB␣ has also been proposed to directly or indirectly affect the p50 NLS accessibility and function, thus limiting nuclear import of NF-B/IB␣ complexes (43,58).
In the present study, we show that the N-terminal domain of IB␣ contains a sequence responsible for the cytoplasmic localization of both IB␣ and pyruvate kinase containing the SV40 large T antigen NLS (PK-NLS). When fused to the glucocorticoid receptor (GR), this motif also prevents the hormone-induced nuclear translocation of GR, indicating its ability to promote cytoplasmic retention of a reporter protein. This sequence, called a cytoplasmic retention sequence (CRS), is recognized by G3BP2, a cytoplasmic protein whose function was so far unknown. G3BP2 not only interacts with IB␣ but also with IB␣/NF-B complexes. G3BP2 is composed by an N-terminal domain homologous to the nuclear transport factor 2 (NTF2), followed by an acidic domain; a domain containing five PXXP motifs, suggesting that G3BP2 could interact with SH3-containing proteins; and a C-terminal domain containing different RNA-binding motifs. The acidic domain is sufficient for the interaction of G3BP2 with the IB␣ CRS, although an additional binding site likely exists. Overexpression of G3BP2 directly promotes retention of IB␣ in the cytoplasm, indicating that subcellular distribution of IB␣ and NF-B/IB␣ complexes likely results from a equilibrium between nuclear import, nuclear export, and cytoplasmic retention.

EXPERIMENTAL PROCEDURES
Antibodies-Rabbit polyclonal antibodies to IB␣, IB␤, and IB⑀ were either purchased from Santa Cruz Biotechnology or kindly provided by Weil et al. (59). Rabbit polyclonal antibodies anti-p65 (C-20) were from Santa Cruz Biotechnology. Mouse monoclonal antibody anti-GST was provided by J. L. Teillaud. The previously described anti-SV5 tag monoclonal antibody (60) was obtained from Dr. R. E. Randall. The anti-Myc tag is the murine monoclonal antibody 9E10. Anti-GFP antibodies were purchased from Roche Molecular Biochemicals and corresponded to a mixture of two mouse monoclonal antibodies (clones 7.1 and 13.1). Mouse monoclonal antibody anti-GR was provided by D. DeFranco. Anti-G3BP2 is a polyclonal rabbit antiserum raised against a peptide encompassing amino acids 225-245 of murine G3BP2 and antigen affinity purified. This antibody is specific for G3BP2 and does not recognize G3BP1 (data not shown).
Plasmids-cDNAs encoding for Myc-tagged versions of PK or PK fused the SV40 NLS region (NLS-PK) were cloned into the HindIII/ KpnI restriction sites of the pcDNA3 vector (61). The indicated fragments of IB␣ were amplified by polymerase chain reaction using the pcDNA3-IB␣ ctag vector (39) as template and cloned into the KpnI/ BamHI restriction sites of NLS-PK encoding vector. Deletion mutants of IB␣ were obtained by amplification of corresponding cDNAs by polymerase chain reaction using the pcDNA3-IB␣ ctag vector (39) as template and cloned into the BamHI/XbaI restriction sites of pcDNA3 (In Vitrogen).
The pSTC-TK-GR3-795 expression plasmid encoding for amino acids 4 -795 of rat GR and the IB␣ NES (265-277)-GR were kindly provided by Dr. D. DeFranco (62,63). To construct the IB␣ 37-55-GR chimera, the IB␣ 37-55 fragment containing a BamHI site at each termini was amplified by polymerase chain reaction and cloned in the BamHI site preceding the fourth amino acid of GR in pSTC-TK-GR3-795. cDNA encoding for human G3BP2 was amplified by polymerase chain reaction from a human brain cDNAs library and sequenced, according to the GenBank TM sequence (accession number AF051311). G3BP2 cDNA was subcloned either into BamHI restriction site in pAcHLT-B plasmid for expression in baculovirus or into BglII/XbaI restriction sites in pSV2 plasmid. cDNAs encoding for human G3BP2 or deletion mutants of G3BP2 were amplified by polymerase chain reaction using the pSV2 vector containing the cDNA of human G3BP2 as template and cloned into the BglII/XbaI restriction sites of the eukaryotic expression vector pEGFP-C1 (CLONTECH) or into the BamHI/ XbaI restrictions sites of pcDNA3-Myc-His (In Vitrogen).
Cell Culture and Transfections-HeLa cells were grown in DMEM supplemented with 10% fetal calf serum. HeLa cells were transfected by electroporation as described (39). A total of 10 g of plasmid DNA encoding the chimeric proteins was transfected in 5 ϫ 10 6 HeLa cells. After transfection, cells were seeded in 4 wells of 6 wells plates for immunofluorescence analysis or in 10-cm dishes for immunoprecipitation and incubation continued for 24 h. Cells transfected with GR expression plasmids were grown in DMEM plus 5% charcoal-stripped fetal calf serum and treated for 1 h with 1 M cycloheximide prior hormone addition. For hormone treatments, transfected cells were incubated with 10 Ϫ7 corticosterone for 1 h in phenol red-free DMEM without fetal calf serum. When indicated, hormone was removed by washing cells three times with phenol red-free DMEM and then incubated with this medium for 2 h. Sf9 cells were grown in Grace's medium supplemented with 10% inactivated fetal calf serum and transfected by calcium phosphate precipitation (Pharmingen).
Infection with Vaccinia Virus and Transfection Procedure-HeLa cells were plated on coverslips 24 h before the experiments. After washing in serum-free medium, cells were infected with the vT7 recombinant vaccinia virus (64,65). Infection was carried out for 30 min at 37°C in serum-free medium containing 25 g/ml soybean trypsin inhibitor and 10 mM Hepes, pH 7.2. After removal of the inoculum, cells were cotransfected using DOTAP (Roche Molecular Biochemicals) with pcDNA3-IB␣ ctag vector (1 g) and either empty pcDNA3 (2 g) or pcDNA3-G3BP2 (2 g). Transfection was carried out in serum-free medium in the presence of 10 mM hydroxyurea to inhibit maturation of vaccinia particles.
Immunoprecipitation-Cells were washed twice with Dulbecco's modified Eagle's medium and once with buffer S (115 mM potassium acetate, pH 7.3, 25 mM Hepes, pH 7.4, 2.5 mM MgCl 2 ) at 37°C and then treated with 2 g/ml streptolysin O (SLO) (53) in Buffer S for 5 min at 37°C. SLO supernatant was kept on ice, cells were washed two times with buffer S containing 0.1% Nonidet P-40, and the resulting washing volume was mixed with SLO supernatant. Alternatively, for large scale preparation of cell extracts, cells were lyzed in buffer T (2% Triton X-100, 150 mM NaCl, 2 mM EDTA, 30 mM Tris, pH 8.6). SLO or Triton X-100 extracts were incubated for 10 min at 4°C and centrifuged at 10,000 ϫ g for 10 min. Appropriate antibodies and protein G-Sepharose beads were added to the supernatants and incubated for 5 h at 4°C. Beads were then washed, boiled for 10 min in Laemmli sample buffer, and analyzed by 8 or 10% SDS-PAGE and silver staining (66) or Western blotting revealed with chemiluminescence protein immunoblotting reagents (POD, Roche Molecular Biochemicals or ECL, Amersham Pharmacia Biotech).
Indirect Immunofluorescence Analysis-For indirect immunofluorescence analysis, HeLa cells grown on coverslips were fixed with 3% paraformaldehyde and permeabilized with 0.1% Triton X-100 for 10 min. Alternatively, cells were treated for 5 min at 4°C with 2 g/ml SLO in buffer S, washed in buffer S, permeabilized for 7 min at 37°C, and then fixed with 3% paraformaldehyde. Primary antibodies were applied for 30 min followed by a 30-min incubation with fluorescein isothiocyanate (FITC) or Texas Red-conjugated donkey anti-mouse or anti-rabbit IgG (Jackson). Coverslips were mounted in Mowiol (Hoechst, Frankfurt, Germany). Images were acquired either with a CCD camera (Princeton) or using confocal laser scanning microscopy with a TCS4D or NT-SP confocal microscope based on a DM microscope interfaced with a mixed gas argon-krypton laser (Leica Laser Technik). Fluorescence acquisitions were performed with the 488-and 568-nm laser lines to excite FITC and Texas Red dyes, respectively, with a ϫ100 oil immersion PL APO objective. Quantifications were performed using the IP Lab software.
Expression of G3BP2 in Insect Sf9 Cells-cDNA encoding for an His-tagged G3BP2 was subcloned in pAcHLT-B (Pharmingen). Sequences were recombined into Autographa californica nuclear polyhedrosis virus. To generate the recombinant virus, 2 g of transfer vector plasmid was cotransfected with 0.25 g of linearized Baculogold baculovirus (Pharmingen) DNA into Sf9 cells. Transfection was carried out following instructions from the manufacturer (Pharmingen). To produce G3BP2 protein, Sf9 cells were infected at the multiplicity of 1 and harvested 72 h after infection. Cellular pellet was resuspended in 50 mM Hepes, pH 7, 100 mM NaCl, 1% Triton X-100, 10% glycerol and centrifuged. Protein quantification was performed on the supernatant by the Coomassie Plus Protein Assay (Pierce).
In Vitro Binding Assays-1.5 g of GST or 3 g GST-IB␣ fusion proteins were immobilized on 25 l of glutathione-agarose beads and incubated with 500 ng of baculovirus expressed G3BP2 for 1 h at 4°C in binding buffer (20 mM Hepes, pH 7.4, 110 mM potassium acetate, pH 7.4, 2 mM magnesium acetate, 0.5 mM EGTA, 1 mM dithiothreitol, 0.1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride). Beads were then washed three times in binding buffer and boiled in Laemmli sample buffer, and bound proteins were analyzed on 10% SDS-PAGE and Western blotting.

RESULTS
The N-terminal Domain of IB␣ Contains a CRS-It has been clearly shown that the ankyrin repeats of IB␣ are responsible for the nuclear import of IB␣ (54,55). This domain also participates in the interaction with NF-B and masks the NLS of NF-B p65. Such a mutual masking of the sequences responsible for the nuclear import of both proteins contributes to the cytoplasmic retention of the NF-B/IB␣ complex. To determine whether the cytoplasmic localization of IB␣ or NF-B/IB␣ complex could be ensured by an additional mechanism, we analyzed the subcellular localization of wt and mutant forms of IB␣. HeLa cells were therefore transiently transfected with plasmid encoding tagged versions of wild type (wt) and deletion mutants of IB␣, and the localization of the resulting proteins was analyzed by indirect immunofluorescence using an anti-tag antibody. As described previously, wt IB␣ was distributed throughout the nucleus and the cytoplasm ( Fig. 1, top panel). Deletion of the C-terminal amino acids 256 -317 did not affect the localization of the resulting protein (data not shown). In contrast, a mutant form of IB␣ lacking amino acids 1-68 was mainly localized in the nucleus with a small fraction expressed in the cytoplasm (Fig. 1, middle panel). To analyze whether this nuclear expression was significant or just resulted from passive diffusion because of the small size of this IB␣ deletion mutant, IB␣ (68 -317) was expressed as a fusion protein with the GFP. As shown in Fig. 1 (bottom panel), the resulting fusion protein was also mainly expressed in the nucleus.
To determine whether the N-terminal domain of IB␣ was sufficient to promote cytoplasmic retention, residues 1-67 of IB␣ were fused to an artificial nuclear protein consisting of Myc-tagged PK containing the SV40 large T antigen NLS ( Fig.  2A). Subcellular distribution of the resulting protein in transiently transfected HeLa cells was analyzed by indirect immunofluorescence using an anti-Myc antibody. In contrast to NLS-PK, which was exclusively expressed in the nuclear compartment, NLS-PK-IB␣ (1-67) was mainly localized in the cytoplasm, with a significant fraction accumulated at the nuclear envelope, but only a small amount of this protein was found in the nucleus ( Fig. 2B and Table I). This result indicates that the N-terminal domain of IB␣ is sufficient to induce the cytoplasmic localization of an NLS-containing protein.
To precisely define the sequence requirement for cytoplasmic localization, intracellular distribution of fusion proteins between different regions of the IB␣ N-terminal domain and NLS-PK was analyzed. Serines 32 and 36 have been shown to be phosphorylated upon cell stimulation. However, mutation of serines 32 and 36 into alanine or aspartate or deletion of residues 16 -26 (⌬16 -26) did not affect the ability of the Nterminal domain of IB␣ to retain NLS-PK in the cytoplasm ( Table I). Fusion of amino acids 1-26 to NLS-PK did not result in the cytoplasmic localization of the fusion protein. In contrast, when residues 27-67 or 27-55 of IB␣ were fused to NLS-PK, the resulting proteins displayed the same distribution as NLS-PK-IB␣ (1-67) ( Fig. 2B and Table I). The minimal sequence required for cytoplasmic localization was refined to residues 37-55 of IB␣ (Table I). Furthermore, mutation of tyrosine 42 into histidine almost completely abolished the ability of the IB␣ 27-55 sequence to localize NLS-PK in the cytoplasm ( Fig. 2B and Table I).
To analyze more precisely the cytoplasmic localization of NLS-PK when fused to the IB␣ 27-55 sequence, HeLa cells were transiently transfected with plasmids encoding PK or NLS-PK-IB␣  and prior to fixation were treated or not with SLO, a bacterial toxin that permeabilizes cells without affecting the integrity of the nuclear envelope. Cells were then processed for indirect immunofluorescence using an anti-Myc antibody, and the intracellular localization of PK and NLS-PK-IB␣  was analyzed by confocal microscopy. In intact cells, PK was distributed throughout the cytoplasm, but no staining was detectable upon SLO treatment indicating that PK is exclusively cytosolic (Fig. 2C). In contrast, NLS-PK-IB␣ (27-55) displayed a cytoplasmic and nuclear envelope localization in the absence of treatment. When cells were permeabilized with SLO prior to fixation, a fraction of this protein was still detectable in the cytoplasm, and the nuclear envelope staining remained intact and appeared punctuate (Fig. 2C), suggesting an interaction with the nuclear pore complex. Taken together, these results indicate that the sequence located between residues 37 and 55 of IB␣: (i) is able to promote the localization of a nuclear reporter protein on unidentified cytoplasmic structures and on the outer membrane of the nuclear envelope and (ii) does not function when Tyr-42 is mutated into His.
A nuclear export sequence has been recently identified in residues 45-54 of IB␣ (42)(43)(44)(45). We also found that NLS-PK-IB␣ (27-55) relocalized to the nucleus upon treatment of cells with leptomycin B, a cytotoxin that inhibits the interaction between CRM1 and NES, indicating that this NES is functional in our experimental model (data not shown). However, to determine whether the sequence located between residues 37 and 55 of IB␣ functions not only as a nuclear export sequence but also as a cytoplasmic retention motif, we took advantage of the regulated nuclear import of GR. In absence of ligand, GR remains predominantly cytoplasmic in association with proteins that limit its access to nuclear import receptors. Upon hormone treatment, GR is released from these heteromeric complexes and rapidly translocates into nucleus where it strongly binds target sites within chromatin and nuclear matrix. Withdrawal of hormone disengages GR from chromatin, but unliganded receptor remains in the nucleus for a long period of time before being exported to the cytoplasm (62). When the HIV-1 Rev protein was fused to the hormone-binding region of GR, the chimeric protein was completely imported into the nucleus and nucleolus upon hormone treatment, indicating that fusion of a very efficient NES to GR is not sufficient to delocalize hormone-bound GR to the cytoplasm. In contrast, hormone withdrawal initiated a very efficient export of the chimeric protein (67). More recently, Liu and DeFranco (63) showed that a chimeric GR containing the IB␣ C-terminal NES displays a rapid and leptomycin B-sensitive nuclear export following hormone withdrawal. We therefore constructed a chimera protein consisting of residues 37-55 of IB␣ fused to amino acid 4 at the N terminus of the rat GR (IB␣ 37-55-GR) and compared the subcellular localization of GR, IB␣ 37-55-GR, and IB␣ NES (265-277)-GR in transiently transfected HeLa cells analyzed by indirect immunofluorescence using anti-rat GR antibodies. In the absence of hormone (Fig. 3, ϪH), GR was predominantly cytoplasmic with a fraction expressed in the nucleus, whereas IB␣ 37-55-GR and IB␣ NES (265-277)-GR were detected only in the cytoplasm. After 30 min of treatment with corticosterone, both GR and IB␣ NES (265-277)-GR exclusively localized in the nuclear compartment as described previously ( (63) ; Fig. 3, ϩH). In contrast, a large fraction of IB␣ 37-55-GR remained cytoplasmic under the same experimental condition. To further confirm that this effect was due to cytoplasmic retention rather than highly efficient nuclear export, distribution of the different chimeric proteins was analyzed after hormone removal. Withdrawal of hormone for 2 h led to a significant nuclear export of IB␣ NES (265-277)-GR, confirming that the NES sequence was functional (Fig. 3, ϩHϩ2h chase, and Ref. 63), although the Rev NES is able to induce a complete nuclear export in the same experimental condition (67). In contrast, hormone removal did not significantly affect the subcellular distribution of GR nor IB␣ 37-55-GR (Fig. 3, ϩHϩ2h chase), indicating that the A, schematic representation of the Myc-tagged fusion proteins, PK, NLS-PK, and NLS-PK coupled to different sequences of IB␣. B, HeLa cells transfected with vectors encoding for the indicated fusion proteins were processed for indirect immunofluorescence and stained with an anti-Myc monoclonal antibody. Primary antibody was detected with a FITCconjugated anti-mouse antibody. Cells were visualized by epifluorescence microscopy, and images were acquired using a CCD camera. C, HeLa cells were transfected with cDNAs encoding for PK or NLS-PK-IB␣ . Cells were permeabilized with Triton X-100 (left panels) after fixation or with SLO prior fixation (right panels) and processed for indirect immuno-fluorescence with an anti-Myc monoclonal antibody followed by a FITC-conjugated anti-mouse antibody. Cells were visualized with a confocal laser scanning microscope, and photographs correspond to the accumulation of four optical sections in one projection. protein PK fused to an NLS and sequences from IB␣ HeLa cells transfected with vectors encoding for the indicated fusion proteins (left column) were processed for indirect immunofluorescence and stained with an anti-Myc monoclonal antibody followed by an FITC-conjugated anti-mouse antibody as described in the legend to Fig.  2. Cells were visualized by epifluorescence microscopy, and intracellular localization of the corresponding proteins is indicated in the right column. N, nucleus; C, cytoplasm; NE, nuclear envelope. Lowercase letters correspond to compartments containing a minor fraction of the protein. CϩNEϩn NLS-PK-IB␣ (47)(48)(49)(50)(51)(52)(53)(54)(55) N sequence located between residues 37 and 55 of IB␣ is not an NES strong enough to delocalize hormone-bound GR to the cytoplasm. Taken together, these data indicate that the sequence located between residues 37 and 55 of IB␣ not only contains an NES but also acts as a CRS.
The Cytoplasmic Retention Motif of IB␣ Interacts with G3BP2-To determine the cellular protein that interacts with IB␣ CRS, HeLa cells were transiently transfected with plasmids encoding Myc-tagged versions of PK, NLS-PK, NLS-PK-IB␣ , or NLS-PK-IB␣(27-55 Y42H), and corresponding cell extracts were immunoprecipitated with an anti-Myc antibody. Immunoprecipitates were then analyzed on SDS-PAGE and silver staining. A 62-kDa protein was found to co-immunoprecipitate with NLS-PK-IB␣  and to coimmunoprecipitate weakly with NLS-PK-IB␣(27-55 Y42H) but was not detected in PK or NLS-PK immunoprecipitates (Fig. 4A). This result suggest that the 62-kDa protein specifically interacts with IB␣ CRS. The 62-kDa band was excised from the gel and digested in situ with Lys-C, and peptides were analyzed by matrix assisted laser desorption and ionization peptide mass mapping. This led to the identification of the 62-kDa species as G3BP2 (AF051311), a protein homologous to G3BP (rasGAP SH3-binding protein; Ref. 68) that we renamed G3BP1. G3BP1 and G3BP2 proteins present 59% identity and display an identical overall molecular organization with an N-terminal domain homologous to the NTF2 protein, a protein involved in nucleocytoplasmic transport (NTF2-like domain; 27% identity and 50% homology with NTF2), followed by an acidic domain (composed of 40% acidic residues), a domain containing five PXXP motifs in G3BP2 and one in G3BP1 (PXXP domain) and a C-terminal domain containing another PXXP motif, RNP2, and RNP1 consensus sequences as well as an RGG-rich region that is therefore related to an RNA-binding domain.
To analyze whether IB␣ interacts with G3BP2 in vitro, purified recombinant G3BP2 was produced in baculovirus-in-fected insect cells and tested for its ability to bind immobilized GST-IB␣. Bound proteins were analyzed by SDS-PAGE and Western blotting using specific anti-peptide antibody against G3BP2 and anti-GST antibody. As shown in Fig. 4B, G3BP2 did not bind to glutathione-agarose beads or GST-coupled beads but bound strongly to immobilized GST-IB␣. Thus a complex between IB␣ and G3BP2 can be formed in vitro.
To determine whether G3BP2 and IB␣ interact in vivo, HeLa cell lysates were immunoprecipitated with anti-G3BP2, anti-IB␣, anti-NF-B p65, or irrelevant antibodies. Immunoprecipitates were analyzed by Western blotting using anti-G3BP2, anti-IB␣, or anti-NF-B p65 antibodies (Fig. 4C). No G3BP2, IB␣, or p65 were detected in mock immunoprecipitate. In contrast, G3BP2 was detected in significant amounts in both IB␣ and p65 immunoprecipitates, demonstrating that G3BP2/IB␣ and G3BP2/IB␣/p65 complexes are formed in vivo. IB␣ and p65 were not found in G3BP2 immunoprecipi- tates. This could be explained either by the weak ability of the anti-G3BP2 antibody to immunoprecipitate, by the high level of endogenous G3BP2 compared with IB␣, or by the ability of the anti-G3BP2 antibody to dissociate G3BP2/IB␣ interaction. Indeed, the anti-G3BP2 antibody is directed against a peptide sequence (residues 225-245) involved in the interaction with IB␣ (Fig. 5); thus it may only precipitate uncomplexed material.
To confirm the interaction between G3BP2 and IB␣ in vivo, HeLa cells were transfected with a plasmid encoding a fusion protein between GFP and G3BP2 (Fig. 5A), cell lysates were immunoprecipitated using anti-IB␣ or anti-GFP antibodies followed by Western blotting with anti-IB␣ or anti-GFP antibodies. As shown in Fig. 4D, GFP-G3BP2 co-precipitated with IB␣ in GFP-G3BP2 transfected cells and not in mock transfected cells, indicating that overexpressed G3BP2 is able to interact with endogenous IB␣. Endogenous IB␣ was very weakly and not reproducibly detectable in GFP-G3BP2 immunoprecipitates. This is likely explained by the high level of GFP-G3BP2 compared with endogenous IB␣. Together, these data indicate that IB␣ or IB␣/p65 complex interact with G3BP2 both in vivo and in vitro through the IB␣ CRS.
To determine whether G3BP2 is able to interact with other members of the IB family, lysates from HeLa cells transfected with a plasmid encoding for GFP-G3BP2 were immunoprecipitated using anti-GFP, anti-IB␣, anti-IB␤, or anti-IB⑀ antibodies followed by Western blotting with anti-GFP or anti-IB antibodies ( Fig. 4E and not shown). GFP-G3BP2 co-precipitated with IB␣ and to a lower extent with IB␤, which are well expressed in HeLa cells. In contrast, expression of IB⑀ was not detectable in HeLa cells (not shown), thus explaining why GFP-G3BP2 was not immunoprecipitated with anti-IB⑀ antibodies. This result indicate that G3BP2 can not only interact with IB␣ but also directly or indirectly with IB␤.
Domains of G3BP2 Responsible for the Interaction with the IB␣ CRS-To analyze the site of binding of G3BP2 to IB␣, a series of deletion mutants of G3BP2 fused to GFP were constructed (Fig. 5A). Upon transfection of the corresponding plasmids in HeLa cells, all these proteins were efficiently immunoprecipitated by the anti-GFP antibody (Fig. 5B, left panel). Immunoprecipitation with an anti-IB␣ antibody showed that deletion of the RNA-binding domain of G3BP2 (GFP-G3BP2 1-299), NTF2-like domain (GFP-G3BP2 117-449), or both (GFP-G3BP2 117-299) did not affect the ability of resulting proteins to interact with IB␣ (Fig. 5B, right panel, lanes 2, 3,  and 4, respectively). This indicates that RNA-binding domain and NTF2-like domain of G3BP2 are not essential for the interaction with IB␣. In contrast, acidic and PXXP domains of G3BP2 are sufficient to mediate binding to IB␣. More precisely, a fusion protein between GFP and the acidic domain of G3BP2 (GFP-G3BP2 117-223) still interacted with IB␣ (Fig.  5B, right panel, lane 6), indicating that the acidic domain of G3BP2 is sufficient for the interaction with IB␣. Although neither the PXXP domain nor the RNA-binding motif separately fused to GFP (GFP-G3BP2 224 -299 and GFP-G3BP2 295-449) were able to coprecipitate with IB␣ (Fig. 5B, right  (68 -317). Cells extracts were immunoprecipitated with anti-SV5 or anti-GFP antibodies, and immunoblotting was performed with anti-SV5 or anti-GFP antibodies as indicated on the left. Note that the stronger interaction of IB␣(1-256) protein with G3BP2 compared with the wt IB␣ was not reproducible and that both proteins interact equivalently with G3BP2. E, indicated GFP-G3BP2 fusion proteins were overexpressed in HeLa cells and localized directly with GFP fluorescence. Cells were visualized with a confocal laser scanning microscope, and photographs correspond to the accumulation of four optical sections in one projection. panel, lanes 7 and 8), fusion of both domains to GFP promoted interaction with IB␣ (Fig. 5B, right panel, lane 5). This suggest that an additional binding site of IB␣ is present in these domains.
G3BP2 Promotes Cytoplasmic Retention of IB␣-Fusion proteins between GFP and wt or deletion mutants of G3BP2 were localized using direct GFP fluorescence and confocal microscopy (Fig. 5E). GFP-G3BP2 (1-449) corresponding to the wt G3BP2 fused to GFP was expressed throughout the cytoplasm and at the nuclear envelope. This localization was unaffected upon treatment with leptomycin B, indicating that CRM1-mediated nuclear export is not involved in G3BP2 distribution (data not shown). The same intracellular localization was observed for the mutant lacking the RNA-binding domain (GFP-G3BP2 1-299) with a small fraction detectable in the nucleus. In contrast, mutant lacking the NTF2-like domain GFP-G3BP2(117-449) displayed a cytoplasmic localization but was not detected to the nuclear envelope. This result indicates that the NTF2-like domain likely targets G3BP2 at the nuclear envelope. It has been shown that the NTF2 protein interacts with some nucleoporins. The NTF2-like domain of G3BP2 might conserve this ability to associate with the nuclear pore complex. The mutant lacking both the RNA-binding and NTF2like domains GFP-G3BP2(117-299) was distributed throughout the nucleus and the cytoplasm but excluded from the nucleoli. Thus, RNA-binding domain and NTF2-like domain appear to be able to retain G3BP2 in the cytoplasm and at the nuclear envelope.
To confirm that G3BP2 participates in the cytoplasmic retention of IB␣, the effect of overexpressed G3BP2 on the intracellular localization of IB␣ was investigated. The G3BP2 protein is abundant in HeLa cells, and classical transfection procedures did not lead to an significant overexpression of the exogenous protein compared with the endogenous level (data not shown). To overexpress G3BP2 protein with respect to its endogenous level, HeLa cells were infected with a T7 recombinant vaccinia virus prior transfection with cDNAs encoding for G3BP2 or IB␣ proteins under the control of T7 promoter. Intracellular distribution of both Myc-tagged G3BP2 and IB␣ was subsequently analyzed by indirect immunofluorescence using anti-Myc and anti-IB␣ antibodies. The ratio of mean of fluorescence intensities between the cytoplasm and the cell nucleus was quantified in each cell using the IP lab software. Using this approach, IB␣ was found mainly localized in the nucleus (Fig. 6, panel a). In contrast, overexpression of G3BP2 led to a significant decrease of the IB␣ nuclear content and concomitant increase in the cytoplasmic level of this protein ( Fig. 6, panels c and e and bottom panel). To analyze the specificity of this result, we examined the effect of G3BP2 overexpression on the localization of an IB␣ protein lacking the N-terminal domain (68 -317). As shown on Fig. 6 (panels b,  d, and f and bottom panel), overexpression of G3BP2 did not affect the subcellular distribution of IB␣ (68 -317) that was mainly expressed in the nuclear compartment. This result indicates that overexpression of G3BP2 led to the retention of IB␣ in the cytoplasm.

DISCUSSION
Regulation of IB␣ turn-over and subcellular localization constitute key checkpoints for modulating NF-B transcriptional activity. The nucleocytoplasmic distribution of IB␣ results from an equilibrium between nuclear import and export. IB␣ contains leucine-rich nuclear export sequences that mediate translocation of both IB␣ and IB␣/NF-B complexes out of the nucleus by a CRM1-mediated transport pathway (39 -45, 48). Nuclear import of IB␣ is mediated by the ankyrin repeats, which are also involved in the interaction with NF-B (54, 55). As a consequence, when IB␣ is not bound to NF-B, it translocates to the nucleus. Conversely, degradation of IB␣ allows NF-B to be imported in the nucleus. However, the crystal structure of NF-B complexed to the IB␣ ankyrin repeat domain reveals that repeats 1 and 2 of IB␣ contact and mask the p65 NLS, whereas the p50 NLS remains surfaceexposed and theoretically accessible for the nuclear import machinery (57). This observation raises the question of the molecular basis responsible for the cytoplasmic localization of the IB␣/NF-B complex at steady state. Two distinct but not exclusive mechanisms have been proposed to account for the cytoplasmic retention of IB␣/NF-B complexes: (i) an active nuclear export mediated by the N-terminal NES promotes continuous export from the nucleus and (ii) the N-terminal region of IB␣ is thought to mask the p50 NLS either directly or indirectly. These dual functions of the N-terminal region are mediated by the same sequence located between residues 45 and 54, but differences are observed in the nature of residues essential for each function (42)(43)(44)58). Data reported here indicate that the same N-terminal domain (residues 37-55) displays properties of a CRS that promotes anchoring of IB␣ in the cytoplasm and on the outer membrane of the nuclear envelope. This CRS constitutes a transferable motif that is sufficient to prevent nuclear import of a protein containing the prototypical basic NLS or an inducible NLS. In the present report, the G3BP2 protein was shown to specifically interact with IB␣ CRS and retain IB␣ in the cytoplasm. This CRS is so far unique to IB␣, but one can note some similarities between IB␣ CRS and the C-terminal domain of murine IB␤ and human IB␤1, which are also able to translocate to the nucleus (69). Indeed, an interaction between G3BP2 and IB␤ was also observed, suggesting that both IB␣ and IB␤ present the ability to be retained in the cytoplasm, and thus confirming the functional redundancy of these two proteins (70). Based on these results, we can speculate that interaction between G3BP2 and IB␣ CRS could regulate the activity of the p50 NLS either by maintaining the N-terminal region of IB␣ in a conformation that directly masks this NLS or by sterically interfering with the interaction of the p50 NLS with the nuclear import machinery. The function of the IB␣ CRS is greatly reduced when Tyr-42 is mutated into histidine. This effect could be explained by a drastic change in the structure of the CRS because of this nonconservative mutation. Alternatively, it has been reported that Tyr-42 is phosphorylated in response to hypoxia, reoxygenation, and pervanadate. This phosphorylation event does not lead to the degradation of IB␣ but results in the dissociation of IB␣ from NF-B, which therefore becomes activated (71)(72)(73). The tyrosine-phosphorylated IB␣ has been shown recently to specifically interact with phosphoinositide 3-kinase, and phosphoinositide 3-kinase activity is required for activation of NF-B in response to pervanadate (74). Therefore, phosphorylation state of Tyr-42 might control interaction of IB␣ with different partners that either prevent IB␣/NF-B complex formation or retain IB␣ or IB␣/NF-B complex in the cytoplasm.
G3BP2, whose function was so far unknown, has been isolated by sequence homology with G3BP1, a protein able to interact with the SH3 domains of the Ras-GTPase activating protein (68,75). Both proteins present 59% identity with 82% identity (96.5% similarity) in their N-terminal domain homologous to the NTF2 protein, 46% identity in their acidic domain (68% similarity), 56% identity in the region containing PXXP motifs (75% similarity), and 64% identity (75% similarity) in the C-terminal domain. This latter domain contains RNP1 and RNP2 consensus sequences as well as an RGG-rich region and is therefore related to an RNA-binding domain. Acidic domain of G3BP2 is sufficient to mediate binding to IB␣, but an additional binding site for IB␣ is likely present within the PXXP-containing domain or the RNA-binding domain. Despite the high degree of identity between G3BP1 and G3BP2, there is no convincing evidence that G3BP1 could bind IB␣ or that G3BP2 could bind RasGAP. 2 Besides the role of G3BP2 in the control of nucleocytoplasmic distribution of IB␣ and cytoplasmic anchoring of the IB␣/NF-B complex, a direct function of G3BP2 in the context of NF-B signaling cascade is so far unclear. Overexpression of G3BP2 did not affect the TNFinduced transcriptional activity of NF-B possibly because of the high level of endogenous G3BP2 (data not shown). On the other hand, G3BP2 could have been involved in targetting IB␣/NF-B complex to the IKK signalsome, a multisubunit complex containing IB␣ kinases as well as structural components NEMO and IKAP in response to cell stimulation (76,77). However, we failed to detect G3BP2 in the IKK signalsome immunopurified using anti-NEMO antibodies in unstimulated or tumor necrosis factor-treated cells (data not shown). Molecular organization of G3BP1 and G3BP2 rather suggest that these protein could correspond to scaffold proteins connecting signal transduction pathways through interaction of their acidic and PXXP-containing domains with RasGAP or IB␣ to RNA metabolism and nuclear transport through their RNAbinding domain and NTF2-like domain, respectively.
The present data indicate that the NTF2-like domain of G3BP2 is likely responsible for targeting G3BP2 to the nuclear envelope and thus might present, like NTF2, the ability to associate with the nuclear pore complex. NTF2 has been reported to play a key role in nuclear transport. Indeed, NTF2 not only interacts with nucleoporins located near the central gated channel (78,79) but also specifically binds the small GTPase Ran in its GDP bound form (80,81). NTF2 facilitates accumulation of Ran in the nucleus where the Ran exchange factor, RCC1, converts RanGDP into RanGTP (82,83). Expression of RCC1 and Ran GTPase-activating protein exclusively in the nucleus and cytoplasm, respectively, creates a gradient of Ran across the nuclear envelope, with RanGDP in the cytoplasm and RanGTP in the nucleus. This gradient is determinant for the directionality of nuclear transport (84). NTF2 is therefore essential to maintain appropriate concentration of Ran across the nuclear envelope. Whether the NTF2-like domain of G3BP2 could bind Ran and eventually facilitate its nuclear import remains to be determined. On the other hand, a role of G3BP2 in the control of the RanGDP concentration in the cytoplasm could be envisaged. Alternatively, C-terminal domains of both G3BP1 and G3BP2 contain RNP1 and RNP2 consensus sequences as well as an RGG-rich region, which are landmarks of hnRNPs, suggesting that G3BP might play a role in mRNA transport. In particular, these protein might be involved in the release of mRNA from the transport complex and reassociation of the exported mRNA with cytoplasmic RNA-binding partners. Interaction of G3BP with signaling proteins might facilitate transport of specific mRNA or alternatively routing specific mRNA to distinct cytoplasmic areas.
G3BP1 has been reported to harbor a phosphorylation-dependent RNase activity responsible for the cleavage of the 3Ј-untranslated region of human c-Myc mRNA, suggesting a connection between the Ras signaling pathway and mRNA decay (85). Partially purified G3BP1 also displays helicase activity that can unwind DNA/DNA, DNA/RNA, and RNA/RNA duplexes (86). Although such activities remain to be analyzed for G3BP2, homologies between G3BP1 and G3BP2 suggest that G3BP2 likely binds RNA. Our data indicate that the RNA-binding domain contributes to the cytoplasmic localization of G3BP2, suggesting that G3BP2 might be bound to RNA-containing structures within the cytoplasm. G3BP2 might therefore influence stability or translational efficiency of particular mRNAs, and such function might be modulated when G3BP2 is associated to IB␣. One could even speculate that this complex might exert a feedback control on IB␣ mRNA metabolism. Identification of specific RNA and protein partners of G3BP2 should allow better understanding of the precise function of this scaffold protein and its link to NF-B transduction pathway.