Regulation of Nuclear Import and Export of Negative Cofactor 2*

The negative cofactor 2 (NC2) is a protein complex composed of two subunits, NC2α and NC2β, and plays a key role in transcription regulation. Here we investigate whether each subunit contains a nuclear localization signal (NLS) that permits individual crossing of the nuclear membrane or whether nuclear import of NC2α and NC2β depends on heterodimerization. Our results from in vitro binding studies and transfection experiments in cultured cells show that each subunit contains a classical NLS (cNLS) that is recognized by the importin α/β heterodimer. Regardless of the individual cNLSs the two NC2 subunits are translocated as a preassembled complex as co-transfection experiments with wild-type and cNLS-deficient NC2 subunits demonstrate. Ran-dependent binding of the nuclear export receptor Crm1/exportin 1 confirmed the presence of a leucine-rich nuclear export signal (NES) in NC2β. In contrast, NC2α does not exhibit a NES. Our results from interspecies heterokaryon assays suggest that heterodimerization with NC2α masks the NES in NC2β, which prevents nuclear export of the NC2 complex. A mutation in either one of the two cNLSs decreases the extent of importin α/β-mediated nuclear import of the NC2 complex. In addition, the NC2 complex can enter the nucleus via a second pathway, facilitated by importin 13. Because importin 13 binds exclusively to the NC2 complex but not to the individual subunits this alternative import pathway depends on sequence elements distributed among the two subunits.

The negative cofactor 2 (NC2) 2 is a protein complex composed of two subunits, NC2␣ (DRAP1) and NC2␤ (Dr1). Both subunits are conserved in eukaryotes and essential for Saccharomyces cerevisae viability (1,2). NC2␣ and NC2␤ heterodimerize via histone-fold domains and associate with the promotor-bound TATA-binding protein (3,4). The resulting NC2-TATA-binding protein-DNA complex sterically hinders the recruitment of transcription factor IIB and in part of transcription factor IIA (5), and thus inhibits transcription initiation (6,7). The NC2 complex is present on a substantial fraction of human genes (8). Besides mediating TATA-binding protein binding to TATA-containing and TATA-less promoters (9) the NC2 complex can also mobilize TATA-binding protein on the DNA (10). In addition to the well established function as transcriptional repressor (11) several studies have shown that NC2 activates transcription, in vitro and in vivo (12)(13)(14)(15)(16). The mechanism underlying the positive effects of NC2 on gene expression is not understood. Although the two NC2 subunits mostly function together, recent studies in S. cerevisae (17), Drosophila (18), and human provided evidence that NC2␣ and NC2␤ can associate with different proteins (19 -21). In this study, we have analyzed whether human NC2 subunits contain localization signals that permit individual crossing of the nuclear membrane or whether nuclear import of NC2␣ and NC2␤ depends on heterodimerization.
During interphase the exclusive site of nucleocytoplasmic exchange is the nuclear pore complex (NPC) (22). Although small molecules can traverse the NPC via diffusion the passage of molecules larger than 40 kDa is restricted by the permeability barrier of the NPC (23). Recent work by Frey and Görlich (24,25) provides evidence that the permeability barrier of the NPC consists of a hydrophobic meshwork with hydrogel-like properties. The passage of this physical barrier requires soluble nuclear transport receptors (also referred to as karyopherins) that recognize intrinsic signal elements displayed on transport cargoes (26). Import signals named nuclear localization signals (NLSs) can be categorized in classical (cNLS) and non-classical (ncNLS) types. Non-classical NLSs are directly recognized by import receptors, whereas the binding of cNLSs requires an additional importin ␣ adapter protein (27). The most abundant nuclear export signals (NESs) are leucine-rich and interact with the export receptor chromosome region maintenance 1 (Crm1), also known as exportin 1 (28,29). A steep RanGTP gradient across the nuclear membrane controls the binding and release of transport cargoes (30).
Here we demonstrate that the NC2 subunits are imported into the nucleus as a preassembled complex. Nuclear accumulation of the NC2 complex occurs via two alternative pathways, facilitated by the importin ␣/␤ heterodimer and importin 13.
The cNLSs present in each subunit have a cumulative effect on the nuclear targeting efficiency of the NC2 complex. In addition to its cNLS the NC2␤ subunit exhibits also a leucine-rich NES, which is masked upon heterodimerization with NC2␣.

EXPERIMENTAL PROCEDURES
Cell Culture-NIH-3T3 (mouse embryonic fibroblast) cells obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ number ACC59) and HeLa P4 cells (31) were cultured in Dulbeccos modified Eagle's medium (Invitrogen). Medium was supplemented with 10% (v/v) fetal bovine serum (Biochrom), antibiotics, and 2 mM glutamine. Cells were maintained in a humidified incubator with 5% CO 2 atmosphere at 37°C.
Expression Constructs-The coding regions of the respective genes were amplified from plasmid DNA using specific primer pairs with appropriate restriction sites.
The bacterial expression constructs were cloned as follows: the coding regions of human NC2␣ and NC2␤ as SpeI/ HindIII fragments into the respective sites of pET-41a(ϩ) (Novagen); the coding regions of human NC2␣ and NC2␤ as NcoI/HindIII fragments into the respective sites of pETM-30 (EMBL, Heidelberg); the coding regions of human NC2␣ and NC2␤ as NdeI/BamHI fragments together with six histidine residues as NcoI/NdeI fragments into the NcoI/ BamHI sites of pET-11d (Novagen); and the coding region of murine Ubc9 as an EcoRI/XhoI fragment into the respective sites of pGEX-4T-1 (GE Healthcare).
The eukaryotic expression constructs were cloned as follows: the coding regions or gene fragments of human NC2␣ and NC2␤ as SalI/BamHI fragments into the respective sites of pEGFP-C1 (Clontech) and pPW1 (modified pEGFP-C1 in which EGFP was replaced by RFP using the restriction sites NheI/BglII); the coding regions of human NC2␣ and NC2␤ as BglII/PstI fragments into the respective sites of pEGFP-EGFP-N1 (modified pEGFP-N1 (Clontech) in which a second EGFP was inserted C-terminal of the MCS as SalI/BamHI fragment); the coding regions of human NC2␣ and NC2␤ as BglII/ SalI fragments from the pEGFP-EGFP-N1 expression constructs into the BglII/SalI sites of pmRFP-N1 (modified pEGFP-N1 in which EGFP was replaced by mono-RFP using restriction sites AgeI/NotI); the gene fragments of NC2␣ (1-10 amino acids) and NC2␤ (97-106 amino acids) as SalI/BamHI fragments into the respective sites of pEGFP-EGFP-GST-C1 (modified pEGFP-C1 in which GST was inserted N-terminal of the MCS as the BglII/XhoI fragment and in which a second EGFP was inserted at the N terminus as NheI fragment); the coding region and gene fragments of human importin 13 as EcoRI/XhoI fragments were inserted into the respective sites of pCS2flag (modified pCS2plus (32,33) in which a FLAG tag was inserted N-terminal of the MCS as a NcoI/EcoRI fragment); the coding regions of human importin ␤, Xenopus importin 7, and murine importin 9 as NruI fragments were inserted into the StuI sites of pCS2flag; and the coding region of human importin 5 as the StuI/XhoI fragment was inserted into the respective sites of pCS2flag. All constructs were verified by DNA sequencing (Andreas Nolte, Abteilung Entwicklungsbiochemie, Universität Göttingen, Germany).
Transfection Experiments-Transfection into HeLa P4 cells was performed with the Effectene TM Transfection Reagent (Qiagen) according to the manufacturer's instructions. The cells were fixed 24 h after transfection with 3% paraformaldehyde in phosphate-buffered saline for 15 min and either analyzed directly by fluorescence microscopy or subjected to indirect immunostaining first. For that purpose, fixed cells were permeabilized with 0.5% Triton X-100 in phosphate-buffered saline for 10 min, blocked with 3% bovine serum albumin in phosphate-buffered saline, and an anti-FLAG polyclonal (rabbit) antibody (Sigma) and anti-rabbit Alexa 488 antibody or anti-rabbit Alexa 555 antibody (Molecular Probes) used to detect FLAG-importin 13. After washing, cells were stained with 10 g/ml Hoechst 33258 (Molecular Probes) and mounted in Histogel (Linaris Histogel).
Heterokaryon Assays-Interspecies heterokaryons of human HeLa P4 cells and mouse NIH-3T3 cells were formed as described previously (34). Briefly, HeLa P4 cells were transiently (co-)transfected with plasmid DNA encoding either fluorescently labeled NC2␣ and NC2␤ or GFP-QKI-5. Thirty hours post-transfection, non-transfected mouse NIH-3T3 cells were co-plated with the HeLa P4 cells and co-cultured for 18 h. The co-culture was then incubated for 2 h in the presence of 50 g/ml cycloheximide and another 30 min in 100 g/ml cycloheximide. The co-cultured cells were washed with phosphatebuffered saline and fused with Sigma HybriMax warmed to 23°C for 2 min. Fused cells were washed with phosphate-buffered saline and incubated at 37°C in medium containing 100 g/ml cycloheximide for an additional 6 h. The cells were fixed with 3% paraformaldehyde in phosphate-buffered saline for 15 min and counterstained with Hoechst 33258 to distinguish between human and mouse cell nuclei.
In Vitro Transcription and Translation-Transcription and translation of human importin 13 fragments (see also Fig. 6A) were performed from the corresponding SP6 promoter constructs (pCS2flag) in vitro. Using the TNT coupled reticulocyte lysate system (Promega) according to the manufacturer's instructions the proteins were labeled with [ 35 S]methionine (Amersham Biosciences). Reactions were performed at 30°C for 2 h in a 12.5-l volume and the samples were then directly used for GST pulldown assays.
Recombinant Protein Expression and Purification-Epitopetagged NC2 complexes were generated as follows: NC2␣ and NC2␤ were co-expressed in Escherichia coli BL21(DE3). The cultures were grown at 37°C to an optical density of 1.0 at 600 nm. After shifting the temperature to 25°C bacterial protein expression was induced with 0.4 mM isopropyl ␤-D-thiogalactopyranoside and the cultures were grown for 4 h. The collected bacteria were resuspended in buffer A (50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 5 mM ␤-mercaptoethanol), lysed by sonication, and the recombinant NC2 complexes were purified on nickel nitrilotriacetic acid-agarose (Qiagen) followed by either gel filtration on Superdex 200 (GE Healthcare) or a second purification step on glutathione-Sepharose 4B (GE Healthcare).
GST Pulldown Assays-GST fusion proteins (or complexes containing GST fusions) immobilized on glutathione-Sepharose 4B were used as affinity matrix for binding experiments. Appropriate amounts of affinity matrix were incubated for 3 h at 4°C with bacterial lysates containing expressed import receptors in buffer B (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 5 mM MgCl 2 , and 5 mM ␤-mercaptoethanol). The binding experiments were performed in the absence or presence of 2 M RanQ69L(GTP). After washing three times with ice-cold buffer B, the affinity matrix was boiled in SDS-PAGE sample buffer and the matrix-bound proteins were analyzed by SDS-PAGE followed by Coomassie staining.
For binding to importin 13 fragments, purified GST-NC2␣/ His 6 -NC2␤ and GST-UBC9 were immobilized on glutathione-Sepharose 4B that had been preincubated with 10% bovine serum albumin in buffer C (50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 5 mM MgCl 2 , and 5 mM ␤-mercaptoethanol). The resulting affinity matrix was washed and incubated with 5 l of the TNT coupled reticulocyte lysate containing in vitro transcribed and translated 35 S-labeled importin 13 fragments in 300 l of buffer C supplemented with 3% bovine serum albumin. After 3 h at 4°C, the matrix was washed three times with buffer C, boiled in SDS-PAGE sample buffer, and matrix-bound proteins were analyzed by SDS-PAGE followed by phosphorimaging (Amersham Biosciences).

Both NC2 Subunits Exhibit a Monopartite cNLS-Classical
NLSs are characterized as short stretches enriched in basic amino acids (43). Based on the consensus motifs for cNLSs (44) each NC2 subunit contains one putative monopartite cNLS. This type of signal follows the four-residue consensus K-(K/R)-X-(K/R) (45) with additional sequence requirements up-and downstream (46). The predicted basic stretch in NC2␣ ( 4 KKKK 7 ) is located in the unstructured N-terminal region, whereas the relevant sequence in NC2␤ ( 100 KRRK 103 ) is found in the predicted random coiled C-terminal half of the fourth ␣ helix (5). To functionally characterize the putative signals we first compared the subcellular localization of wild-type and mutated NC2 subunits heterologously expressed in HeLa P4 cells (Fig. 1). To visualize the subcellular distribution, both NC2 subunits were fused to green (EGFP) and red (RFP) fluorescent proteins. Surprisingly, the subcellular distribution of the NC2 subunits showed a strong dependence on the position of the fluorescent fusion protein. Tandem EGFP or RFP fused to the C terminus caused a dominant cytoplasmic distribution of the individual NC2 subunits (supplemental Fig. S1A). In contrast, fusion of EGFP or RFP to the N terminus caused a nuclear localization pattern of NC2␣ and a rather homogeneous distribution of NC2␤ ( Fig. 1 and supplemental Fig. S1B). Because the subcellular distribution of N-terminal fusion proteins was similar to the localization of endogenous NC2 subunits (data not shown) they were used for further analysis. We observed that wild-type RFP-NC2␣ no longer accumulated in the nucleus when either lysine residue 4 (K4A) or lysine residue 5 (K5A) was mutated (Fig. 1A). Similar results were found for NC2␤, the homogeneous distribution of wild-type EGFP-NC2␤ was blocked when either lysine residue 100 (K100A) or arginine residue 101 (R101A) of the putative cNLS was substituted for alanine ( Fig. 1B). Because the absence of neither critical lysine residues at the second position nor basic amino acid residues at the third position of the four-residue consensus sequence were tolerated, both sequences appeared to represent cNLSs. To analyze whether the predicted sequences are sufficient to mediate nuclear transport of a reporter protein, amino acids 1-10 of NC2␣ ( 1 MPSKKKKYNA 10 ) and 97-106 of NC2␤ ( 97 VALKRRKASS 106 ) containing the putative cNLSs (underlined) were fused to EGFP-EGFP-GST (EEG). EEG alone resides exclusively in the cytoplasm when expressed in mammalian cells, because it is too large to enter the nucleus by passive diffusion and does not contain a NLS (Fig. 1C). Fusion of amino acids 1-10 of NC2␣ or amino acids 97-106 of NC2␤ to EEG led to efficient nuclear import of the resulting proteins ( Fig. 1C). To identify nuclear transport receptors that interact with the individual NC2 subunits, in vitro binding studies were performed. GST-tagged NC2␣ and NC2␤ were expressed in E. coli and immobilized on glutathione-Sepharose. The immobilized fusion proteins were incubated with the importin ␣/␤ heterodimer, importin ␤, transportin 1, importin 5, importin 7, and importin 13, all from bacterial lysates (Fig. 1D). After washing, the bound proteins were analyzed by SDS-PAGE followed by Coomassie staining. The importin ␣/␤ heterodimer was bound to both NC2 subunits. In the nucleus direct binding of RanGTP to ␤-family import receptors disintegrates the receptor-substrate interaction. The binding of importin ␣/␤ to the NC2 subunits was abolished in the presence of RanGTP, demonstrating its specificity (Fig. 1D). Additionally, weak binding of transportin 1 was observed. However, this interaction was not RanGTP-dependent (data not shown). None of the other import receptors, including importin ␤ alone, bound significantly to the individual NC2 subunits. Together, these results demonstrate that each subunit exhibits a monopartite cNLS, 4 KKKK 7 in NC2␣ and 100 KRRK 103 in NC2␤, which are necessary and sufficient for nuclear uptake of the respective NC2 subunit.
is not homogeneously distributed like EGFP alone but strictly cytoplasmically, which can only be explained by the presence of an NES within the first 100 amino acids of NC2␤. The most abundant and also the best-characterized NESs are leucine-rich (enriched in hydrophobic amino acids) and interact with the export receptor Crm 1/exportin 1 (28,29). Thus, we examined whether the cytoplasmic localization of fluorescently labeled NC2␤ was lost on the application of leptomycin B (LMB), a specific inhibitor of exportin 1 (28). We observed that wild-type EGFP-NC2␤ and NC2␤-RFP largely accumulated in the nucleus on LMB treatment for 2 h (Fig. 2, A and B). In contrast to NC2␤, the strong cytoplasmic distribution of C-terminallabeled NC2␣ was not affected by LMB treatment (data not shown) and N-terminal-labeled NC2␣ showed a largely nuclear localization in transfected HeLa P4 cells (see Fig. 1A and supplemental Fig. S1B). Hence, these results do not provide evidence for the existence of an NES in NC2␣. As LMB impairs various cellular pathways the effects of LMB were additionally verified. The sequence analysis of the first 100 amino acids of NC2␤ revealed one sequence element ( 71 VIQALESLGF 80 ) that fulfills the loosely defined consensus motif for leucine-rich NES: This putative NES was initially confirmed by the fact that amino acids 63-88 of NC2␤ fused to EGFP (EGFP-NC2␤-(63-88)) localized more cytoplasmically than EGFP alone. Furthermore, nuclear export of NC2␤ was strongly reduced in cultured cells when the last two hydrophobic amino acids of the putative NES were mutated ( 71 VIQALESLGF 80 -71 VIQALESAGA 80 ; mutated residues are underlined), as shown for the EGFP-NC2␤-(L78A/ F80A) and NC2␤-RFP-(L78A/F80A) fusion proteins (Fig. 2, A  and B). The results of the mutational analysis resemble the effect of LMB treatment and indicate that the mutated amino acids are part of a hydrophobic NES recognized by exportin 1. The direct interaction of exportin 1 and NC2␤ was verified by in vitro binding studies. In contrast to import receptors RanGTP is required for the binding of cargoes to export receptors. Recombinant GST-NC2␤ bound specifically to exportin 1 in the presence of RanGTP, whereas no efficient binding was observed in the absence of RanGTP (Fig. 2C). The RanGTP-dependent binding of exportin 1 to GST-NC2␤-(L78A/F80A), however, was reduced. Thus, hydrophobic amino acids leucine (Leu 78 ) and phenylalanine (Phe 80 ) are necessary for the interaction of exportin 1 and NC2␤. Together, these results demonstrate that NC2␤ con- tains a leucine-rich NES ( 71 VIQALESLGF 80 ), which confers nuclear export via exportin 1. The location of the mapped cNLSs and NES are summarized in supplemental Fig. S2.
Dimerization with NC2␣ Masks the NES in NC2␤-Many transcription factors contain both types of signals, NLS and NES, which allows them to shuttle between the nucleus and the cytoplasm (49). The identification of the leucine-rich NES in NC2␤ raised the question whether also the NC2 complex, which is nuclear at steady state (Fig. 3A), can be exported. To answer this question, we performed interspecies heterokaryon assays (34,50). HeLa P4 cells were transiently co-transfected with plasmid DNA encoding wild-type NC2␣ and NC2␤ fused to EGFP-EGFP or RFP (Fig. 3, A and B). The fusion of fluorescent proteins to the C terminus of NC2 subunits seems to favor nuclear export as indicated by the cytoplasmic localization of heterologously expressed NC2 subunits (supplemental Fig.  S1A). However, the C-terminal fusions do not block nuclear import in general as shown by the nuclear accumulation of NC2␤-RFP in cells treated with LMB (Fig. 2B), and the predominantly nuclear distribution of co-expressed NC2␣ and NC2␤ subunits (Fig. 3A). Twenty-four hours post-transfection, an equal amount of untransfected mouse NIH-3T3 cells were seeded to the transfected HeLa P4 cells, the cells were co-cultured, and ultimately fused to form heterokaryons. To inhibit new protein synthesis cycloheximide was added and to distinguish between human HeLa P4 and mouse NIH-3T3 nuclei the cells were stained with Hoechst. Six hours after cell fusion, colocalized NC2␣ and NC2␤ were detected in only 5% of the mouse nuclei (white arrows) of heterokaryotic cells (Fig. 3, B  and D). This indicates that the NES in NC2␤ does not allow rapid shuttling of the NC2 complex. RanGTP-dependent binding of exportin 1 to GST-NC2␤ could no longer be detected when the subunit was dimerized with His 6 -NC2␣. Instead, unspecific binding of exportin 1 to the GST-NC2␤/His 6 -NC2␣ complex in the absence of RanGTP was observed (supplemental Fig. S3). The shuttling protein QUAKING-5 fused to GFP (GFP-QKI-5) was used as a positive control. In all heterokaryons, GFP-QKI-5 accumulated in the nucleus of non-transfected mouse NIH-3T3 cells (Fig. 3, C and D) demonstrating that QKI-5 shuttles between the nucleus and cytoplasm as previously described (51).
Both cNLSs Are Necessary for Nuclear Import of the NC2 Complex-Next we examined whether nuclear accumulation of the NC2 complex relies on just one or both of the cNLSs of its subunits. For that purpose, we co-expressed the wild-type and cNLS-mutated NC2 subunits described above (see again Fig. 1,  A and B) in HeLa P4 cells. Because the heterologously expressed NC2 complex is mainly comprised of fluorescently tagged components, the individual contribution of cNLSs of NC2␣ and NC2␤ could be elucidated. The results of these transfection studies are summarized in Fig. 4. Quantitative analysis (Fig. 4B) revealed an 80% nuclear localization of co-expressed wild-type RFP-NC2␣ and EGFP-NC2␤ (Fig. 4A, top panel). Mutations in either the cNLS of NC2␣ (K5A) or the cNLS of NC2␤ (R101A) did not only reduce the nuclear accumulation of the mutant proteins to 60% or less but also of the corresponding wild-type partner subunit, leading to a homogeneous distribution of the NC2 complexes (Fig. 4A, middle panels). The mutual localiza-FIGURE 3. One-way transport of the NC2 complex. HeLa P4 cells were transiently co-transfected with plasmid DNA encoding wild-type NC2␣ and NC2␤ fused to EGFP-EGFP or RFP (A and B). The shuttling protein QUAKING-5 fused to GFP (GFP-QKI-5) was overexpressed in HeLa P4 cells (C). A, the subcellular distribution of the gene products was examined 24 h post-transfection by direct fluorescence. Co-expression of NC2␣ and NC2␤ leads to a nuclear co-localization shown in yellow (merge). The DNA was counterstained with Hoechst. B, 30 h posttransfection, untransfected mouse NIH-3T3 cells were co-seeded onto the HeLa P4 cells, co-cultured for 18 h, and fused with Sigma HybriMaxா. The subcellular distribution of the fusion proteins was examined after an additional 6-h incubation in the presence of cycloheximide. To distinguish between HeLa P4 and NIH-3T3 cell nuclei the heterokaryotic cells were counterstained with Hoechst. Colocalized NC2␣ and NC2␤ (shown in yellow, merge) could hardly be detected in the nucleus of mouse NIH-3T3 cells (white arrows). Thus, the NES in NC2␤ does not allow nucleo-cytoplasmic shuttling of the NC2 complex. C, GFP-QKI-5 accumulates in the nucleus of non-transfected mouse NIH-3T3 cells (white arrow) demonstrating that QKI-5 shuttles between the nucleus and cytoplasm. D, quantification of two independent experiments. 100 heterokaryotic cells were analyzed per condition. Only co-transfected HeLa P4 cells were included in the analysis of the NC2 complex. Bars indicate the mean Ϯ S.D.
tion dependence of the NC2 subunits observed in these experiments suggests that the two subunits assemble essentially in the cytoplasm and are transported into the nucleus as heterodimers. In contrast to the mutation of a single cNLS, cytoplasmic accumulation of the NC2 complex was substantially more pronounced when both cNLSs were absent (Fig. 4A, bottom panel). Nuclear import of the cNLS-deficient RFP-NC2␣/ EGFP-NC2␤ complex dropped to 35% (Fig. 4B). Similar results were obtained with EGFP-NC2␣/RFP-NC2␤ complexes (supplemental Fig. S4). Thus, these results demonstrate that the two cNLSs have a cumulative effect on the nuclear targeting efficiency of the NC2 complex.
The NC2 Complex Also Enters the Nucleus via Importin 13-To identify nuclear transport receptors that directly bind to the NC2 complex, in vitro interaction studies were performed. For that purpose, GST-tagged NC2␣ and His 6 -tagged NC2␤ were co-expressed in E. coli, affinity-purified on nickel-nitrilotriacetic acid, and immobilized on glutathione-Sepharose. The immobilized GST-NC2␣/His 6 -NC2␤ complex was incubated with the importin ␣/␤ heterodimer, importin ␤, transportin 1, importin 5, importin 7, and importin 13, all from bacterial lysates (Fig. 5C). After washing, the bound proteins were analyzed by SDS-PAGE followed by Coomassie staining. Surprisingly, importin 13 was much more efficiently bound to the immobilized NC2 complex than the importin ␣/␤ heterodimer. Both NC2-receptor interactions were reduced in the presence of RanGTP suggesting functional significance. None of the other transport receptors bound significantly to the NC2 complex. Because of the strong binding of importin 13 to the NC2 complex in vitro, we tested its possible role as NC2 carrier also in cultured cells. For that purpose, FLAG-tagged importin 13 was additionally co-expressed with mutated RFP-NC2␣ (K5A) and EGFP-NC2␤ (R101A). The co-expression of exogenous importin 13 led to an efficient nuclear accumulation of the cNLS-deficient RFP-NC2␣/EGFP-NC2␤ complex (Fig. 5A). In contrast to importin 13, the additional co-expression of FLAGtagged importin ␤, importin 5, importin 7, and importin 9 did not result in enhanced nuclear uptake of the NC2 complex devoid of cNLSs. Fig. 5B shows a quantitative analysis of the data in Fig. 5A. These results demonstrate that the loss of nuclear import via the classical importin ␣/␤ pathway could only be rescued by importin 13. Comparable results were obtained with the cNLS-deficient EGFP-NC2␣/RFP-NC2␤ complex (supplemental Fig. S5). Hence, a second, alternative, transport pathway via importin 13 does exist for the NC2 complex. However, the subcellular distribution of wild-type and cNLS-mutated RFP-NC2␣ (K5A) was not influenced by co-expression of importin 13 (supplemental Fig. S6A), and the subcellular localization of NC2␤-RFP was only slightly changed upon co-expression of importin 13 (supplemental Fig. S6B). These results are in line with previous in vitro binding studies showing that importin 13 does not interact with the monomeric NC2 subunits (see Fig. 1D). In conclusion, importin 13 mediates nuclear import of the NC2 complex but not of its individual subunits. In that way, the NC2 complex behaves like its closest relative, the NF-YB/NF-YC histone-fold heterodimer. Only the heterodimerized NF-YB/NF-YC subunits were specifically recognized by importin 13, as shown previously (52). Because heterodimerization of both NC2␣ with NC2␤ and NF-YB with NF-YC is a prerequisite for importin 13 binding, the minimal sequence elements recognized by importin 13 comprise the histone-fold domains of both NC2 subunits. Moreover, basic amino acid residues distributed among the two subunits and . Nuclear accumulation of the NC2 complex via importin ␣/␤ requires the cNLSs of both subunits. A, HeLa P4 cells were transiently cotransfected with plasmid DNA encoding wild-type (wt) and mutated RFP-NC2␣ and EGFP-NC2␤, respectively. The subcellular distribution of the gene products was examined 24 h post-transfection by direct fluorescence. The DNA was counterstained with Hoechst. Co-expression of wild-type RFP-NC2␣ and EGFP-NC2␤ results in a nuclear co-localization shown in yellow (merge; top panel). Mutation of either the cNLS of NC2␣ (K5A) or the cNLS of NC2␤ (R101A) reduced the nuclear accumulation leading to a homogeneous localization of both subunits (middle panels). Nuclear import of the RFP-NC2␣/ EGFP-NC2␤ complex was strongly blocked when the cNLSs of both subunits were mutated (bottom panel). B, for quantitative analysis, the mean red (RFP) and green (EGFP) fluorescence value in the nucleus and cytoplasm of 25 cells that co-expressed RFP-NC2␣ and EGFP-NC2␤ was measured using the ImageJ Software (NIH). After subtraction of the background value the percentage of nuclear localization of the different NC2 complexes was calculated. Bars indicate the mean Ϯ S.D.
conserved between related histone-fold pairs (Lys 18 , Arg 19 , Arg 40 , and Lys 63 in NC2␣ and Lys 64 , Lys 88 , Lys 95 , and Arg 102 in NC2␤) are crucial for the recognition of importin 13. 3 The C Terminus of Importin 13 Is Dispensable for Recognition of the NC2 Complex-After confirming that importin 13 indeed represents an alternative transport receptor for nuclear import of the NC2 complex, we wanted to characterize the binding sites in importin 13 for the NC2 complex. Additionally, we compared the binding of NC2 with that of another importin 13 substrate, the sumo-conjugating enzyme UBC9 (40). For that purpose, we generated different importin 13 deletion constructs and tested their NC2 binding potentials (Fig. 6). Independent of the HEAT repeat prediction for importin 13 in Swiss-Prot, the secondary structure prediction programs PHD (Heidelberg) and PSIPRED (53,54) were used to identify putative ␣ helices within importin 13. The programs predicted 38 and 37 ␣ helices, respectively, which is in agreement with the assumption that all transport receptors of the importin ␤ family share 19 HEAT repeats (55). The deletion constructs were created by cutting between the predicted ␣ helices (putative HEAT repeats) of importin 13, and the importin 13 fragments were in vitro transcribed and translated (Fig. 6, A and B). The immobilized GST-NC2␣/His 6 -NC2␤ complex and GST-UBC9 were incubated with the 35 S-labeled importin 13 fragments and bound fractions were analyzed by SDS-PAGE followed by phosphorimaging. As shown in Fig. 6B, among the importin 13 fragments only amino acids 1-669 and 1-784 of importin 13 were bound to the NC2 complex (middle panel) and UBC9 (lower panel). Importin 13 fragments lacking the N terminus did not significantly bind to the immobilized substrates. Accordingly, only C-terminal truncations of importin 13 (amino acids 1-784 FIGURE 5. Importin 13 also mediates nuclear transport of the NC2 complex. A, HeLa P4 cells were transiently co-transfected with plasmid DNA encoding mutated RFP-NC2␣ (K5A), EGFP-NC2␤ (R101A), and FLAG-tagged import receptors. ⌻he subcellular distribution of the gene products was examined 24 h post-transfection by direct fluorescence of the RFP and EGFP fusion proteins. The overlap between the green EGFP fusion protein and the red RFP fusion protein is shown in yellow (merge). The DNA was counterstained with Hoechst. RFP-NC2␣-(K5A) and EGFP-NC2␤-(R101A) both predominantly remained in the cytoplasm of co-transfected cells. Although co-expression of importin 13 led to a strong nuclear accumulation of the NC2 complex, co-expressed importin ␤, importin 5, importin 7, and importin 9 did not affect the subcellular distribution of the NC2 complex. B, quantification of nuclear import of the cNLS-deficient NC2 complex. The mean fluorescence value of co-localized RFP-NC2␣-(K5A) and EGFP-NC2␤-(R101A) was measured in 15 cells using ImageJ software (NIH). The percentage of nuclear localization in the presence of co-expressed import receptors was calculated. Bars indicate the mean Ϯ S.D. C, immobilized GST-NC2␣/ His 6 -NC2␤ complex was incubated with bacterial lysates containing the indicated import receptors. Bound fractions were analyzed by SDS-PAGE and Coomassie stained. The NC2 complex binds strongly to importin 13 and less pronounced to the importin ␣/␤ heterodimer. Both cargo-receptor interactions are reduced in the presence of RanGTP, which was used to simulate nuclear conditions. Mw, molecular weight; imp, importin; trn, transportin 1. and 1-669, respectively) were able to efficiently accumulate the cNLS-deficient RFP-NC2␣/EGFP-NC2␤ complex in the nucleus of co-transfected HeLa cells (supplemental Fig. S7). Due to a similar binding pattern of the importin 13 fragments to the NC2 complex and UBC9, we also asked whether the binding sites in importin 13 for the NC2 complex and UBC9 overlap. To address this question binding of recombinant purified importin 13 to an immobilized GST-NC2␣/His 6 -NC2␤ complex was performed in the absence and presence of equal amounts of recombinant, untagged UBC9 (Fig. 6C). The importin 13 binding was reduced in the presence of UBC9, which demonstrates that the binding sites in importin 13 for the NC2 complex and UBC9 overlap.
In summary, our results suggest that the two NC2 subunits preassemble in the cytoplasm prior to nuclear import. The NC2 complex enters the nucleus via two alternative pathways, facilitated either by the importin ␣/␤ heterodimer or importin 13. Furthermore, each import pathway depends on sequence elements distributed on both NC2 subunits.

DISCUSSION
In this study, we have analyzed the nuclear import pathways of the human NC2 complex. We identified a cNLS in each NC2 subunit. Because cNLSs bind in an extended conformation to importin ␣ they are most likely located in unstructured regions to fit into the importin ␣ binding pockets (44,56). The cNLS of FIGURE 6. The C terminus of importin 13 is dispensable for recognition of the NC2 complex. A, the names of the importin 13 expression constructs used and the amino acids contained in the constructs are listed. Lines indicate the deleted regions and gray bars represent the different importin 13 fragments. B, the minimal importin 13 fragment that binds to the NC2 complex and UBC9 consists of amino acids 1-669. The two NC2 subunits, GST-NC2␣ and His 6 -NC2␤, as well as GST-UBC9 were (co-)expressed in E. coli and used as bait after immobilization on glutathione-Sepharose. The immobilized GST-NC2␣/His 6 -NC2␤ complex and GST-UBC9 were incubated with in vitro transcribed and translated 35 S-labeled importin 13 fragments; all from the TNT coupled reticulocyte lysate. Starting material (20% of the importin 13 fragments that were used) and bound fractions were analyzed by SDS-PAGE followed by phosphorimaging (Amersham Biosciences). Among the importin 13 fragments, wild-type importin 13, and amino acids 1-669 of importin 13 showed the highest binding competence for the NC2 complex (see relative binding in percent below the gels, quantified with the program ImageQuant 5.2). A similar binding pattern was observed for UBC9; also here importin 13 fragments lacking the N terminus did not significantly bind to the immobilized substrate. C, the binding sites in importin 13 for the NC2 complex and UBC9 overlap. Importin 13 interacts with the NC2 complex in an UBC9 sensitive manner. Importin 13 was bound specifically to the immobilized GST-NC2␣/His 6 -NC2␤ complex and this binding was reduced in the presence of equal amounts of UBC9. MW, molecular mass in kilodalton; imp, importin; aa, amino acids. NC2␣ ( 4 KKKK 7 ) is located in the N terminus, whereas the cNLS of NC2␤ ( 100 KRRK 103 ) is found in the center of the protein. In the native proteins, these regions are predicted to be unstructured or random coiled, respectively (5). In addition, both cNLSs fulfill the consensus motif for a monopartite cNLS as proposed by Chelsky et al. (45). However, none of the additional sequence requirements for monopartite cNLSs summarized by Conti (46) are satisfied in the two NC2 subunits. Differences from preferred amino acids for monopartite cNLSs may account for the weak interaction of the importin ␣/␤ heterodimer with the NC2 subunits in vitro. Interestingly, the cNLS of NC2␣ with four contiguous lysines resembles the downstream cluster of the bipartite cNLS of nucleoplasmin. Perhaps puzzling, whereas the downstream cluster of nucleoplasmin is not sufficient to target a protein into the nucleus (57,58), amino acids 1-10 of NC2␣ containing its cNLS ( 4 KKKK 7 ) functioned as an effective monopartite signal when fused to an EGFP-EGFP-GST reporter protein. Both NC2 genes are highly conserved in eukaryotes and are required for viability in yeast (1). However, the monopartite cNLS of neither NC2␣ nor NC2␤ is conserved in S. cerevisiae. Therefore, nuclear import of NC2 in yeast must occur via a different transport mechanism.
Despite exposing cNLSs the two human NC2 subunits are not imported individually but assemble essentially in the cytoplasm and are transported into the nucleus as heterodimer. This was demonstrated by transfection experiments in which the loss of the cNLS in one subunit strongly affected the nuclear accumulation of the partner subunit. Nuclear import of the NC2 complex was significantly blocked when the cNLSs of both subunits were mutated. These results point to an additive effect of both cNLSs on the nuclear targeting efficiency of the NC2 complex.
Besides the cNLS also a leucine-rich NES ( 71 VIQALESLGF 80 ) was identified in the NC2␤ subunit. In S. cerevisiae the last of the four essential hydrophobic positions (underlined) contains tyrosine instead of phenylalanine. However, in contrast to the human subunit three alternative putative sequence elements fulfill the consensus motif for a leucine-rich NES in yeast. The presence of NLS and NES allow NC2␤ to shuttle between the cytoplasm and nucleus. Phosphorylation near an NLS can either have a stimulatory or inhibitory function for protein transport (59). For example, it has been shown that phosphorylation of residues in close proximity to the NLS of the transcriptional activator NF-AT (60) and Lamin B2 (61) lead to the inhibition of the NLS function. The NC2␤ subunit is phosphorylated by casein kinase II in vitro and has two serine residues (Ser) in close distance to the cNLS ( 100 KRRKASS 106 ) (3). Substitution of the two serines for glutamate (S105E/S106E) to mimic the effect of phosphorylation did not significantly influence nuclear accumulation of the NC2 complex (supplemental Fig. S8A). In contrast, nuclear import of NC2␤-(S105E/S106E) fused to EGFP-EGFP-GST was reduced in cells treated with LMB (supplemental Fig. S8B) indicating that the cNLS function can be inhibited by phosphorylation. Hence, the modulation of nuclear transport rates by cytoplasmic phosphorylation may represent a mechanism for regulating nuclear activity of NC2␤.
The results of interspecies heterokaryon assays with fluorescently labeled NC2 subunits demonstrated that the NC2 complex does not shuttle between the nucleus and cytoplasm.
Hence, heterodimerization with NC2␣ masks the NES of NC2␤ and ensures that only complexed NC2␤ remains in the nucleus. These findings point toward a scenario in which NC2␣ also functions independently of NC2␤. More detailed, the activity of the NC2 complex could be modulated by displacement of NC2␤. Due to the immediate nuclear export of removed NC2␤ interactions of NC2␣ with additional nuclear partners such as BTAF1 (19) would not be disrupted.
The NES in NC2␤ is located at the C terminus of the third ␣ helix (5). This ␣ helix is part of the histone-fold motif that facilitates dimerization of the two NC2 subunits. Based on the structural information provided by Kamada et al. (5), only the valine residue at position 71 is involved in subunit dimerization, whereas none of the other essential hydrophobic residues of the NES are predicted to be involved. Therefore, we assume that binding of exportin 1 to the NC2 complex is blocked via steric hindrance. Likewise, Stommel et al. (62) showed that tetramerization masks the NES of p53 and prevents its export.
Surprisingly, endogenous importin 13 alone was not sufficient to facilitate effective nuclear import of the heterologously expressed NC2 complex. One would expect that strong overexpression of importin 13 cargoes can be compensated by constant recycling of the transport receptor. However, very low cellular concentrations of endogenous importin 13 could explain the cytoplasmic retention of the overexpressed NC2 complex. Additionally, the affinity of importin 13 to the NC2 complex might be low in vivo due to phosphorylation of the NC2 subunits (3). As recently described for its closest relative, the NF-YB/NF-YC histone-fold heterodimer, only the dimerized NC2 subunits were recognized by importin 13 (52). Hence, despite their functional divergence these two histone-fold heterodimers rely on the same import mechanism to translocate through the NPC. Interestingly, for importin 13 the most similar protein from S. cerevisiae is Pdr6p/karyopherin 122, which translocates the NC2 antagonist transcription factor IIA into the nucleus (68). The identity of Pdr6p with importin 13 is too low to regard it as a clear orthologue (40). However, Pdr6p might be the functional homologue of importin 13 in yeast. Hence, it would be interesting to investigate whether the yeast NC2 complex (Ydr1-Bur6) is transported by Pdr6p and whether human transcription factor IIA can be imported by importin 13.
To characterize the binding sites in importin 13 for the NC2 complex we compared the binding of several importin 13 fragments. The results of these in vitro interaction studies demonstrated that the C terminus of importin 13 is dispensable for recognition of the NC2 complex. These results were further confirmed by co-transfection experiments with the importin 13 fragments. In contrast to the observations made by Tao and colleagues (69) all FLAG-tagged importin 13 fragments including the N-terminal truncated fragments distributed equally between the cytoplasm and nucleus. 4 The loss of the N terminus of importin 13 did therefore not result in a dominant negative regulator. Despite the homogenous distribution, none of the N-terminal-truncated importin 13 fragment facilitated efficient nuclear uptake of the NC2 complex. Hence, the N terminus of importin 13 is essential for recognition and nuclear import of the NC2 complex. Furthermore, we showed that importin 13 cannot bind the NC2 complex and UBC9 at the same time due to overlapping importin 13 binding sites.
For importin ␤, two major types of cargo-receptor interactions are found. These interactions are either hydrophobic as in the case of the SREBP-2 NLS where minimal contacts are made between importin ␤ and the substrate (70) or electrostatic as observed for the importin ␤ binding domain of importin ␣ (71). Because dimerization of the two NC2 subunits buries the hydrophobic residues of the second ␣ helix and creates a positively charged surface (5) it is reasonable to assume that importin 13 and NC2 interact via electrostatic interactions. The requirement of basic amino acid residues for importin 13-mediated nuclear import of the NC2 complex 3 provides first support for this hypothesis.