Importin alpha nuclear localization signal binding sites for STAT1, STAT2, and influenza A virus nucleoprotein.

Proteins actively transported into the nucleus via the classical nuclear import pathway contain nuclear localization signals (NLSs), which are recognized by the family of importin alpha molecules. Importin alpha contains 10 armadillo (arm) repeats, of which the N-terminal arm repeats 2-4 have been considered as the "major" NLS binding site. Interferon-activated, dimerized signal transducers and activators of transcription (STAT1 and STAT2) directly bind to importin alpha5 via a dimeric nonclassical NLS. Here we show by site-directed mutagenesis that the very C-terminal arm repeats 8 and 9 of importin alpha5 form a unique binding site for STAT1 homodimers and STAT1-STAT2 heterodimers. Influenza A virus nucleoprotein also contains a nonclassical NLS that is recognized by the C-terminal NLS binding site of importin alpha5, comprising arm repeats 7-9. Binding of influenza A virus nucleoprotein to importin alpha3 also occurs via the C-terminal arm repeats. Simian virus 40 large T antigen instead binds to the major N-terminal arm repeats of importin alpha3, indicating that one importin alpha molecule is able to use either its N- or C-terminal arm repeats for binding various NLS containing proteins.

Regulated import of molecules into the nucleus through the nuclear pores is a vital event in eukaryotic cells. Importins (also known as karyopherins) are the major cargo carriers from the cytoplasm into the nucleus. Large molecules (Ͼ40 kDa) that cannot passively diffuse through the nuclear pores use a signal-mediated transport system. The importin ␣/importin ␤-mediated import pathway was the first one to be discovered, and it is also referred to as the classical pathway. Proteins transported into the nucleus contain nuclear location signals (NLSs) 1 that are recognized by importin ␣/importin ␤ heterodimers. Importin ␣ recognizes and binds the NLS, and importin ␤ docks the complex to the nuclear pore and translocates it into the nucleus (1-3).
The crystal structures of two importin ␣ molecules, yeast importin ␣ (10) and mouse importin ␣2 (11), have been determined. Importin ␣ is composed of a large central domain that consist of 10 tandemly repeated armadillo (arm) motifs, which are organized in a superhelix flanked by small N-and Cterminal domains. The 10 arm repeats generate an array of binding pockets that are situated within a long helical surface groove. One binding pocket typically includes a tryptophan residue, followed by an asparagine residue 4 residues downstream. The arm repeats are variable within one protein, but they are remarkably conserved in sequence and order when homologous proteins from yeast to humans are compared. The N-terminal importin ␤ binding domain of importin ␣ mediates binding to importin ␤ (12,13). In the absence of cargo and importin ␤, the importin ␤ binding domain blocks the NLS binding site of importin ␣ (14,15). The C-terminal domain of importin ␣ mediates interactions with the export receptor CAS (16,17).
The NLS used by the classical nuclear import pathway is a short stretch of positively charged amino acids, arginines and lysines, that lack strict consensus sequence (18). A monopartite NLS, like that of simian virus 40 (SV40) large T antigen, is composed of a cluster of five to seven basic amino acids (19,20). A typical bipartite NLS contains two clusters of basic amino acids separated by a linker of 10 -11 amino acids (21).
Monopartite NLSs have been shown to bind to the "major" binding site of importin ␣, which is formed by arm repeats 2-4 (10). The downstream cluster of the bipartite NLS is also recognized by these arm repeats, whereas the upstream cluster is recognized by arm repeats 7 and 8, also called the "minor" binding site (11).
Signal transducers and activators of transcription (STATs) are latent cytoplasmic transcription factors that regulate the expression of a number of genes involved in host defense and growth (22). Binding of interferon (IFN) to its specific cellsurface receptor leads to the activation of receptor-associated Janus (Jak) tyrosine kinases, which phosphorylate STATs. This leads to dimerization and translocation of STATs into the nucleus. Dimerization is an essential and sufficient step for nuclear import (23). Although bound by importin ␣5, STATs do not contain a classical NLS (24). Instead, it is supposed that both STAT1 and STAT2 contain in the DNA binding domain a * This work was supported by the Medical Research Council of the Academy of Finland and by the Sigrid Juselius and the Finnish Cancer Foundations. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed: National Public Health Institute, Mannerheimintie 166, FIN-00300 Helsinki, Finland. Tel.: 358-9-47448377; Fax: 358-9-47448355; E-mail: krister.melen@ ktl.fi. 1 The abbreviations used are: NLSs, nuclear localization signals; NP, nucleoprotein; STAT, signal transducers and activators of transcription; IFN, interferon; GST, glutathione S-transferase; BSA, bovine serum albumin. nonclassical NLS that seems to become operative in STAT dimers (25)(26)(27). In response to IFN-␥ stimulation, STAT1 homodimers are formed, whereas IFN-␣ stimulation results in the formation of STAT1-STAT2 heterodimers. Both STAT1 homodimers and STAT1-STAT2 heterodimers are specifically bound to importin ␣5 (24,26,28). The STAT1-binding site of importin ␣5 has been suggested to be located in the very Cterminal end of STAT1 apart from the arm repeat domain (24). Also other regions or completely different mechanisms have been suspected to affect the nuclear translocation of STATs. Strehlow and Schindler (29) suggested that the most N-terminal amino acids (the first 129 residues) regulate the nuclear import of STAT molecules. Subramaniam et al. (30) suggested that the C-terminal nuclear localization sequence of IFN-␥ regulates STAT1␣ nuclear import, and Bild et al. (31) showed that receptor-mediated endocytosis is necessary for STAT3 nuclear import.
Influenza A virus nucleoprotein (NP) and SV40 large T antigen have functioned as important model molecules for analyzing the nuclear import machinery. Influenza A virus NP encapsidates the viral genome and is essential for viral transcription, replication, and packaging. NP interacts with viral RNA, NP, and other viral proteins as well as cellular proteins including importins ␣1 and ␣5 (32, 33). Two NLSs have been identified in this protein, an unusual NLS at the N terminus (34) and a classical bipartite NLS in the middle of the molecule (35). Apparently, the N-terminal NLS mediates NP interaction with importins (34,36). SV40 large T antigen has a monopartite NLS that interacts with importin ␣ (19,20,37).
Here we show by site-directed mutagenesis that importin ␣5 recognizes the nonclassical STAT1, STAT2, and influenza A virus NP NLSs by its arm repeat domain. The binding site for STAT NLS is composed of arm repeats 8 and 9 and for influenza A virus NP of arm repeats 7-9. Both binding sites differ from the previously described major and minor classical NLS binding sites of importin ␣. We also show that influenza A virus NP binds to the C terminus and SV40 large T antigen to the N-terminal NLS binding site of importin ␣3, indicating that one importin ␣ molecule is able to use different binding sites for various NLSs.
Plasmids and DNA Manipulations-The importin ␣1 and ␣3 gene constructs encoding C-terminal His-tagged proteins have been described previously (9). His tags were replaced by GST tags from pFA6aGST-HIS3Mx6 via BamHI/HindIII using PCR mutagenesis. Importin ␤ and ␣5 cDNAs in GST expression vectors were kindly provided by Drs. D. Görlich  Human importin ␣7 gene (9) (GenBank TM accession number AF060543) was PCR-modified with oligonucleotides GAG CGG ATC CAC CAT GGA GAC CAT GGC GAG CCC (5Ј oligonucleotide, initiation codon underlined) and TTC TTA GGA TCC CTA TTA TAG CTG GAA GCC CTC (3Ј oligonucleotide) to create BamHI cloning sites (in boldface) on both sides of the gene coding region. After BamHI digestion the insert was cloned into the pGEX-2T GST fusion vector (Amersham Biosciences). Baculovirus expression constructs of SV40 large T antigen and influenza A virus NP have been described previously (26,41). All DNA manipulations were performed according to standard protocols, and the newly created gene constructs were partially sequenced.
Baculovirus Expression-Human Tyk2 cDNA in the baculovirus expression plasmid pVL1392 and SV40 large T antigen cDNA in the baculovirus expression plasmid were kindly provided by Dr. Sandra Pellegrini (Institute Pasteur, Paris) and Dr. J. M. Pipas (University of Pittsburgh, Pittsburgh (41)), respectively. Wt STAT1 and STAT2 baculovirus constructs were as described (25). Influenza A (PR8) virus NP gene was inserted into the BamHI site of the pAcYM1 expression plasmid, and recombinant viruses were obtained by plaque purification as described (38). For protein production Sf 9 cells were coinfected with Tyk2 and STAT protein-expressing baculoviruses for 42 h. Influenza A NP and SV40 large T antigen were produced without coinfection with Tyk2. Virus-infected Sf 9 cells were collected, and the whole cell extracts were prepared by disrupting the cells in 50 mM Tris-HCl buffer, pH 7.4, 150 mM NaCl, 5 mM EDTA, and 1% Triton X-100 (immunoprecipitation (IP) buffer) on ice for 30 min. The cells were disrupted with a syringe. Cell extracts were clarified by Eppendorf centrifugation (13,000 rpm, 10 min).
Importin Binding Assay, SDS-PAGE, and Western Blotting-Human importins ␤, ␣1, ␣3, ␣5, and ␣7 as well as mutants in the arm repeats 2, 3, 2 ϩ 3, 2 ϩ 4, 3 ϩ 4, 2 ϩ 3 ϩ 4, 2 ϩ 3 ϩ 7, 2 ϩ 3 ϩ 8, 7, 8, and 9 of importin ␣5 and mutants in the arm repeats 3 and 8 of importin ␣3 were expressed in Escherichia coli BL21 cells as GST fusion proteins under isopropyl-1-thio-␤-D-galactopyranoside induction. Bacteria were lysed in IP buffer with 5 mg/ml lysozyme (Sigma) at room temperature for 30 min, briefly sonicated, and clarified by Eppendorf centrifugation (13,000 rpm, 5 min). From 0.1 to 1.0 ml of bacterial cell extracts containing GST-importin proteins were allowed to bind to 25 l of glutathione-Sepharose 4 Fast Flow beads (Amersham Biosciences) in IP buffer with agitation at ϩ4°C for 60 min, followed by washing twice with the buffer. From 0.1 to 1.0 ml of baculovirus cell extracts containing STATs, influenza A NP, or SV40 large T antigen as well as 50 l of in vitro translated influenza A virus NP protein (TNT Coupled Reticulocyte Lysate Systems, Promega, Madison, WI) were allowed to bind in IP buffer to Sepharose-immobilized GST-importins ␣3 and ␣5 or GSTimportin ␤ at ϩ4°C for 2 h, followed by washing 3 times with the buffer. Sepharose beads were dissolved in 50 l of 2ϫ Laemmli sample buffer, and the proteins were separated on 8% SDS-PAGE (42). The gels were stained with Coomassie Brilliant Blue or transferred onto Immobilon-P membranes (polyvinylidine difluoride; Millipore, Bedford, MA), followed by staining with primary and secondary antibodies and visualization of the proteins with the enhanced chemiluminescence system (Amersham Biosciences) as recommended by the manufacturer. 35 S-Labeled protein samples were separated on 8% SDS-PAGE. The gels were stained and treated with Amplify reagent (Amersham Biosciences) as recommended and autoradiographed.

RESULTS
Structural Features of the Conserved Importin ␣ Molecules-Importin ␣ molecules are evolutionarily conserved and can be found in eukaryotes from yeast to humans. In humans, six importin ␣ isoforms have been identified (Fig. 1A). Yeast cells contain only one importin ␣ isoform that shows over 55% se- Cell extracts were prepared, and the proteins in the cell extracts were allowed to bind to Sepharose-immobilized GST-importins ␣1, ␣3, ␣5, or ␣7 at ϩ4°C for 2 h. Unbound proteins were washed, and Sepharose-bound proteins were dissolved in Laemmli sample buffer, separated on 8% SDS-PAGE, transferred to Immobilon membranes, and stained for STAT1 and STAT2 as indicated in the figure. A similar gel was also stained with Coomassie Brilliant Blue to visualize the amount of Sepharose-immobilized GST-importin ␣ isoforms.
quence similarity with the human importins. The similarity of the six human importins is between 56 and 89% (Fig. 1A).
The central domain of importin ␣ is composed of 10 arm repeats (Fig. 1B). The well conserved tryptophan and asparagine residues in arm repeats 2-4 and 7-9 in the helical surface groove of the importin ␣ molecule are thought to be crucial for NLS binding (10, 11) (Fig. 1, B and C). An alignment of the central domain of human importin ␣ molecules is shown in Fig.  1C. Typical arm repeats contain a tryptophan-asparagine pair at analogous positions. A tryptophan-asparagine pair was found in arm repeats 2-4 and 7-8. Arm repeats 5 and 6, which do not contain this pair, have not been shown to participate directly in NLS binding. Arm repeat 9 contains the conserved asparagine residue, but the tryptophan residue is replaced either by an aspartic acid (importins ␣1, ␣3, and ␣4) or an asparagine (importins ␣5, ␣6, and ␣7). Based on published crystal structures (10, 11) and the alignment in Fig. 1C, series of mutations were created to analyze the interactions between importin ␣ and various NLSs. Arm repeat 9 was included in our analysis, because the two asparagine residues appeared as part of a possible NLS binding pocket (Fig. 1C).
IFN-stimulated STAT Dimers Bind Specifically to Importin ␣5-At present, STAT1 homodimers and STAT1-STAT2 heterodimers are the only STAT dimers that have been shown to interact with importin ␣ molecules (24,26,28). Recent mutational analyses suppose that leucine 407 and lysines 410 and 413 within the DNA binding domain of STAT1 function as an NLS and mediate binding to importin ␣5 (25,26,28).
To determine the possible interactions of STAT1-STAT2 dimers with other importin isoforms than ␣5, we stimulated the human NK-92 cell line with 1000 IU/ml of IFN-␣ for 30 min. Cell extracts were prepared, and the cellular proteins were allowed to bind to Sepharose-immobilized GST-importins ␣1, ␣3, ␣5, or ␣7 at ϩ4°C for 2 h, followed by an analysis of the bound STAT proteins by Western blotting. As shown in Fig. 2, STAT1-STAT2 heterodimers were readily bound to importin FIG. 3. Binding of STAT dimers, influenza A virus NP, and SV40 large T antigen to different importin ␣ isoforms. A, Sf 9 cells were infected with Tyk2-and STAT1-expressing (pSTAT lanes) or STAT1-expressing (STAT1 lanes) ecombinant baculoviruses for 42 h. The cells were collected cell extracts prepared, and the soluble proteins were allowed to bind to Sepharose-immobilized GST-importin (GST-imp.) ␣ isoforms at ϩ4°C for 2 h. After washing, importin ␣-bound proteins were dissolved in Laemmli sample buffer, followed by separation of the proteins on 8% SDS-PAGE, transfer to Immobilon membranes, and staining for STAT1. A similar gel was also stained with Coomassie Brilliant Blue to visualize the amount of Sepharose-immobilized GST-importin ␣ isoforms and STAT1. Coomassie Brilliant Blue-stained phosphorylated STAT1 is marked with dots. B, baculovirus-infected Sf 9 cell extracts containing either STAT1 and STAT2 proteins or STAT1, STAT2, and Tyk2 proteins (pSTAT1ϩpSTAT2) were allowed to bind to GST-importin ␣ proteins as described above and stained for STAT1 and STAT2. C, the control Sf 9 cell extracts or Sf 9 cell extracts containing influenza A virus NP protein (influenza A NP) were allowed to interact with GST-importin ␣ subtypes as described above and stained for NP. D, the control Sf 9 cell extract or Sf 9 cell extracts containing SV40 large T antigen (SV40 T Ag) was allowed to interact with GST-importin ␣ isoforms as described above and stained for T antigen.
Tyrosine-phosphorylated STAT Dimers, Influenza A NP, and SV40 Large T Antigen Bind to Different Importin ␣ Isoforms-To characterize further STAT binding to different importin ␣ isoforms, we used a baculovirus expression system to reconstitute STAT activation (26). Coinfection of Sf 9 cells with recombinant Tyk2 and STAT protein expressing baculoviruses resulted in an efficient expression and tyrosine phosphorylation of STAT1 and STAT2 proteins. Tyrosine-phosphorylated STAT1 and STAT2 formed dimers, which enabled us to analyze the potential interactions of dimeric STAT complexes with importins (25,26). As shown in Fig. 3, baculovirus-expressed STAT1 homodimers or STAT1-STAT2 heterodimers bound specifically to Sepharose-immobilized GST-importin ␣5 but not to importin ␣1, ␣3, or ␣7 isoforms. These results, together with those shown in Fig. 2 using natural IFN-␣-stimulated cells, show that the reconstituted STAT activation in the baculovirus expression system is a reliable method to study STAT-importin ␣ interactions.
We also analyzed interactions of Sepharose-immobilized GST-importins ␣1, ␣3, ␣5, or ␣7 with baculovirus-expressed influenza A NP and SV40 large T antigen. Two NLSs have been identified in influenza A virus NP, a nonclassical NLS at the N terminus (34) and a classical bipartite NLS in the middle of the molecule (35). It has been shown that the N-terminal NLS of NP is involved in the interaction with importins (34,36). Here we show that influenza A virus NP bound strongly to importins ␣1 and ␣5 and to a lesser extent to importin ␣3, whereas no marked binding to importin ␣7 isoform was detected (Fig. 3C).
Importins ␣1 and ␣5 have been shown to bind efficiently SV40 large T antigen NLS peptide or T antigen NLS conjugated to BSA (24,(43)(44). Instead, our binding experiments of SV40 large T antigen with Sepharose-immobilized GST-importins revealed that full-length T antigen bound strongly to importin ␣3 and to a lesser extent to importin ␣1 but not to importin ␣5 or ␣7 isoforms (Fig. 3D), indicating that the natural context of an NLS can greatly affect its accessibility to importins.
A C-terminal Unique NLS Binding Site, Including Arm Repeats 8 and 9 of Importin ␣5, Binds STATs-Because STAT NLS is formed by a cluster of basic amino acids within the DNA binding domain of STAT1 and STAT2 (25,27) and recognized by importin ␣5 (26, 28), we reasoned that importin ␣5 arm repeats might be involved in STAT NLS binding. To analyze the role of arm repeats in STAT-importin ␣5 interaction, we created several mutations in the binding pockets of arm repeats 2-4 and 7-9 of importin ␣5 (Figs. 1 and 4).
The C-terminal Arm Repeats of Importin ␣5 Bind Influenza A Virus NP-Because STAT1 homodimers and STAT1-STAT2 heterodimers bound to the C-terminal arm repeats 8 and 9 of importin ␣5, it was of interest to study the binding site for influenza A NP that also has a nonclassical NLS. Our results show that the binding site for influenza A NP, comprising arm repeats 7-9, is also located in the C-terminal half of importin ␣5 (Fig. 5C). To confirm the results found with baculovirusexpressed influenza A NP, we also used in vitro translated NP protein in GST-importin binding experiments (Fig. 5D). Both methods gave quite similar results. A weak binding was detected with arm repeat 9. Arm repeat 7 was not involved in STAT-NLS binding, but it was absolutely necessary for influenza A NP binding. The importin ␣5-binding site for influenza A NP thus seems to be a larger one compared with that of STATs.
Influenza A Virus NP and SV40 Large T Antigen Bind to Different NLS Binding Sites of Importin ␣3-To identify influenza A virus NP and SV40 large T antigen-binding sites of importin ␣3, we mutated arm repeats 3 or 8 of the GSTimportin ␣3 gene construct. Point mutations in the arm repeat 8 of importin ␣3 clearly reduced but did not completely prevent influenza A virus NP binding. Mutations in the arm repeat 3 instead had no effect on NP-importin ␣3 interaction (Fig. 6A). The arm repeat 3 mutant of importin ␣3 (W191A and N195A) had completely lost its ability to bind SV40 large T antigen. Unlike the case of influenza A NP, mutations in the arm repeat 8 (W390A and N394A) had no effect on T antigen binding (Fig.  6B). These results clearly show that one importin ␣ molecule can use either its N or C terminus for binding various NLS containing proteins.

DISCUSSION
Importin ␣ molecules play a central role in the classical nuclear import pathway. In this study we investigated the motives by which importin ␣5 recognizes the nonclassical STAT NLS. The major NLS binding site, comprising arm repeats 2-4 of importin ␣ molecules, is known to bind classical basic type NLS motifs. It has been suggested previously that the importin ␣5-binding site for STAT1 NLS is distinct from the classical NLS binding site and is located in the C terminus (24). However, the exact binding site was not determined, and it was suggested that a new mechanism, not involving any arm repeats, was in use. Here we show by site-directed mutagenesis that the C-terminal arm repeats 8 and 9 mediate the binding of STAT1 homodimers and STAT2 heterodimers to importin ␣5. Bank accession number 1bk5). The representation shows the positions of amino acids that were mutated in importin ␣5 N-terminal (arm repeats 2-4) or C-terminal (arm repeats 7-9) NLS binding domains. Tryptophans are shown in red (also asparagine 445 in arm repeat 9) and asparagines in purple. Two additional amino acids mutated in arm repeats 2 or 7 are shown in green.

FIG. 4. A space-filling representation of an importin molecule (Protein Data
Also the nonclassical NLS of influenza A NP is recognized by the C-terminal arm repeats of importin ␣5. We also show that importin ␣3 uses its N-and C-terminal binding sites for binding various NLS-containing molecules. Crystallographic analyses of importins bound to NLS peptides show that also the C-terminal minor binding site of importin ␣, comprising arm repeats 7 and 8, is able to bind NLS peptides. The major NLS binding site of yeast importin ␣ molecule recognizes the classical monopartite NLS peptide (SV40 large T antigen NLS, PKKKRKV) and the larger downstream basic cluster of the classical bipartite NLS peptide (nucleoplasmin NLS, KRPAATKKAGQAKKKKLD). The minor NLS binding site has a supportive role in binding the smaller upstream basic cluster of the bipartite NLS (10,37). Mouse importin ␣ showed a similar binding mode for the bipartite nucleoplasmin NLS as yeast importin ␣, but SV40 large T antigen NLS peptide was bound almost equally by the major and minor binding sites (11). Also c-myc NLS peptide (PAAKRVKLD) binds both to major and minor binding sites of yeast importin ␣ (37).
The C-terminal half of importin ␣ has been suggested to be involved in NLS binding of several proteins. By using deletion analyses it has been shown that the NLSs of a dimeric DNAbinding protein nuclear antigen 1 of Epstein-Barr virus (KRPRSPSS) and the matrix protein of human immunodeficiency virus (KKKYKL) are bound by the C-terminal half of importin ␣1 (45, 46). The NLS of a human transcription factor FIG. 5. Binding of STATs and influenza A virus NP to arm repeat mutant importin ␣ proteins. A, baculovirus-infected Sf 9 cell extracts containing either STAT1 protein (STAT1) or STAT1 and Tyk2 proteins (pSTAT1) were allowed to bind to Sepharose-immobilized GST-importin ␣5 wt protein or arm repeat mutants of importin ␣5 at ϩ4°C for 2 h as shown in the figure. Binding was performed as described in the legends for Fig. 2. Western blots were stained with anti-STAT1 immunoglobulins. A similar gel was stained with Coomassie Brilliant Blue to visualize the amount of Sepharose-immobilized GST-importin (GST-imp.) ␣ proteins and pSTAT1. Coomassie Brilliant Blue-stained STAT1 is marked with dots. B, an importin ␣5 binding experiment with baculovirus-infected Sf 9 cell extracts containing either STAT1 and STAT2 proteins or STAT1, STAT2, and Tyk2 proteins (pSTAT1ϩpSTAT2) were performed as above and stained for STAT1 and STAT2. C, the control Sf 9 cell extract or Sf 9 cell extract containing influenza A virus NP protein were allowed to bind to importin ␣5 molecules as above and stained for NP. D, in vitro translated influenza A virus NP protein was allowed to bind to Sepharose-immobilized GST, GST-importin ␤, GST-importin ␣5 wt, or arm repeat mutants of importin ␣5 as shown in the figure at ϩ4°C for 2 h. An autography of [ 35 S]methionine-labeled NP is shown.
LEF-1 (NLSKKKKRKREK) is recognized by importins ␣1 and ␣5 and was mapped to the C terminus of importin ␣1 (16). All these NLSs contain a classical monopartite NLS consensus sequence K(K/R)X(K/R) (where X is any residue). Herpesvirus saimiri open reading frame 57 is a multifunctional trans-regulatory protein that interacts via its nonclassical arginine-rich NLS with importins ␣1 and ␣5. By a series of deletion mutants of importins ␣1 and ␣5, it was shown that both importins bind herpesvirus saimiri open reading frame 57 NLS by their Cterminal arm repeats (47). The sequence of the NLS is RRPSR-PFRKP, and the arginine and lysine residues were shown to be essential for importin binding. The C-terminal arm repeats of importins ␣1 and ␣5 thus seem to be able to bind both classical and nonclassical NLSs.
Importin ␣ molecules bind classical NLSs mainly through their basic residues, although the surrounding amino acids also contribute to the binding (10,11,37). It appears that the nonclassical NLSs that are known to bind to importin ␣ molecules contain at least a few basic residues. The potential NLS of STATs is no exception. It contains basic residues but also the adjacent residues are of importance (25,28). Our results, revealing the STAT-binding sites in importin ␣5, further support the idea that a cluster of basic residues in the DNA binding domain of STAT1 and STAT2 forms an essential part of the STAT NLS.
STAT dimers contain two basic elements, one in each STAT monomer, and that both of these elements are required for nuclear import is an extraordinary feature (25,27). One STAT dimer has been shown to bind to two importin ␣ molecules (26). Our hypothesis is that one importin ␣5 molecule is bound to each of the STAT molecules in the dimer, and additional interactions between the two importins or importins and STATs are needed to stabilize the import complex. It has also been shown that large cargoes need more than one importin molecule for a rapid nuclear import to take place (48). A similar dimeric basic type NLS has also been found in immediate early protein (IE1) of baculovirus Autographa californica (49). It has been shown previously that importin ␣5 is necessary for nuclear translocation of phosphorylated STAT, whereas importin ␣1 is not (24). Our data provide further evidence that importin ␣5 is the only isoform mediating nuclear import of STAT1 homodimers, because isoforms ␣1, ␣3, and ␣7 did not bind STAT1. Influenza A NP, on the other hand, is bound by importins ␣1, ␣3, and ␣5. Importins ␣1 and ␣5 are known to bind to the N-terminal nonclassical NLS of influenza A NP (34,36). In the present study we show that importin ␣5 uses its C-terminal arm repeats 7-9 for the binding of NP NLS. Although importin ␣5 binds by its C-terminal arm repeats both STATs and influenza A NP, the binding sites are not identical. STATs need only arm repeats 8 and 9 for binding, whereas influenza A NP-importin ␣5 interaction requires arm repeats 7-9.
In the present study, we show that importin ␣3 is able to use both N-and C-terminal-binding sites for binding different nuclearly targeted proteins. Previous experiments showed that SV40 large T antigen NLS peptides and T antigen NLS coupled to BSA bound to the N-terminal major binding site of importins ␣1 and ␣5 (24). Our studies show that the full-length SV40 large T antigen is bound to the N-terminal major binding site and influenza A virus NP, on the other hand, to the C-terminal binding site of importin ␣3.
Our results suggest that the location and surroundings play an important role in determining the accessibility of an NLS to importins. We show here that the full-length SV40 large T antigen was recognized by importin ␣3 and weakly by importin ␣1. However, as a peptide or as a peptide conjugated to a protein not naturally transported into the nucleus (e.g. BSA), T antigen NLS has been shown to bind to importins ␣1, ␣5, and weakly to ␣3 (24,43,44). The accessibility of an NLS can apparently be regulated by phosphorylation or structural masking (50). SV40 large T antigen is known to have a phosphorylation site affecting nuclear transport (51).

FIG. 6. Binding of influenza A virus NP and SV40 large T antigen to wild type and arm repeat 3 and 8 mutant importin ␣3
proteins. A, the control Sf 9 cell extract or influenza A virus NPexpressing Sf 9 cell extracts were allowed to bind to Sepharose-immobilized wt or arm repeat 3 or 8 mutants of GST-importin ␣3 at ϩ4°C for 2 h. The binding experiment was carried out as described in the legend for Fig. 2. Importin ␣3-bound proteins were separated on 8% SDS-PAGE, transferred to Immobilon filters, and stained for NP. A similar gel was also stained with Coomassie Brilliant Blue to visualize the amount of Sepharose-immobilized GST-importin ␣ molecules. Coomassie Brilliant Blue-stained NP is marked with dots. B, the control Sf 9 cell extract or baculovirus-infected Sf 9 cell extracts containing SV40 large T antigen (SV40 T Ag) were allowed to bind to wt or mutant importin ␣3 molecules as described above, and the proteins bound to beads were processed for Western blotting and stained for T antigen. Coomassie Brilliant Blue-stained T antigen is marked with dots.