Phospholipid Scramblase 1 Contains a Nonclassical Nuclear Localization Signal with Unique Binding Site in Importin α*

Nuclear import of proteins containing a classical nuclear localization signal (NLS) is an energy-dependent process that requires the heterodimer importin α/β. Three to six basic contiguous arginine/lysine residues characterize a classical NLS and are thought to form a basic patch on the surface of the import cargo. In this study, we have characterized the NLS of phospholipid scramblase 1 (PLSCR1), a lipid-binding protein that enters the nucleus via the nonclassical NLS 257GKISKHWTGI266. This import sequence lacks a contiguous stretch of positively charged residues, and it is enriched in hydrophobic residues. We have determined the 2.2 Å crystal structure of a complex between the PLSCR1 NLS and the armadillo repeat core of vertebrate importin α. Our crystallographic analysis reveals that PLSCR1 NLS binds to armadillo repeats 1–4 of importin α, but its interaction partially overlaps the classical NLS binding site. Two PLSCR1 lysines occupy the canonical positions indicated as P2 and P5. Moreover, we present in vivo evidence that the critical lysine at position P2, which is essential in other known NLS sequences, is dispensable in PLSCR1 NLS. Taken together, these data provide insight into a novel nuclear localization signal that presents a distinct motif for binding to importin α.

Nuclear transport is an active signal-mediated process that requires, in most cases, soluble transport factors and specific import signals. Two families of transport receptors have been identified in the importin ␤ superfamily, which is involved in both nuclear import and export, and the TAP superfamily that mediates nuclear export. Transport receptors recognize specific nuclear localization signals (NLSs) 1 and nuclear export signals exposed on the molecular surface of cargoes. In the classical nuclear import pathway proteins bearing a classical SV40-like NLS (PKKKRKV) are recognized by the importin ␣/␤ heterodimer (also known as karyopherin ␣/␤) (1)(2)(3). Importin ␣ (4) acts as an adaptor that recognizes NLS sequences after association with the receptor importin ␤ (5). The importin ␣/␤-NLS cargo complex is then translocated through the nuclear pore complex in a process that requires multiple rounds of interaction of the receptor importin ␤ with nucleoporins, likely via their exposed hydrophobic FG-rich motifs (6,7). Release of the import complex from nuclear pore complex binding sites as well as the final release of the import cargo into the nucleoplasm is mediated by the small GTPase Ran, which binds to importin ␤ in its nuclear GTP-bound form (8).
Structural and biochemical investigations of importins have provided great help in dissecting the molecular determinants for recognition of NLS sequences (9). Importin ␤ presents a modular structure built of 19 tandem HEAT repeats (10). Each HEAT repeat is a simple secondary structural motif formed by two helices (A and B) connected by a short loop (11,12). In the tertiary structure of importin ␤, 19 HEAT repeats are arranged to form a superhelix of helices, which exposes two structurally and functionally distinct surfaces to the solvent (10). The outer surface of the protein contains at least two major nucleoporin binding sites (7,13), whereas the internal face presents an extended cargo binding surface that also binds Ran (8). Three different cargoes have been visualized in complex with importin ␤ using crystallographic techniques: the importin ␤ binding (IBB) domain of importin ␣ which binds to the C-terminal HEAT repeats 7-19 (10); the parathyroid hormone-related protein NLS, which interacts with the N-terminal HEAT repeats 2-11 (14); and the sterol regulatory element-binding protein 2, which binds as a dimer to HEAT repeats 7-17 (15).
The adaptor importin ␣ also consists of two functionally and structurally distinct domains. The highly basic IBB domain is located N-terminally of a helical core built by 10 armadillo (Arm) repeats (16). The C-terminal Arm repeat of the protein also contains an "acidic tail" that is necessary for binding to the export factor CAS/Cse1 (17). The Arm repeat structurally resembles a HEAT repeat and differs from it by being composed of three helices (A, B, and C) connected by two loops (12). The helical Arm core of importin ␣ provides a structurally rigid scaffold that exposes a concave and a convex surface. Two NLS binding sites have been identified in the concave surface of the protein. A major binding site is located between Arm repeats 2 and 4 and a minor binding site, which presents a smaller area of binding, between Arm repeats 7 and 8 (16). Crystallographic analysis of yeast and vertebrate importin ␣ in complex with NLS peptides has revealed that two monopartite NLS peptides bind simultaneously to both major and minor NLS binding sites (18 -20). It is unclear, however, if the binding to the minor NLS binding site is physiologically relevant or reflects a crys-tallization artifact resulting from the large excess of NLS peptide used for co-crystallization with importin ␣. In contrast, a single nucleoplasmin bipartite NLS interacts with both NLS binding sites via two basic boxes, which are spaced apart by ϳ10 nonconserved residues (KRPAATKKAGQAKKKK) (20,21).
In addition to providing a high affinity for importin ␤, the IBB domain of importin ␣ also contains an autoinhibitory NLSlike sequence that regulates binding of cargoes to the adaptor (22,23). When importin ␣ is not bound to importin ␤, the autoinhibitory sequence within the IBB domain occupies the NLS binding pocket, resulting in a low affinity (ϳM) for NLS cargo. Binding of the IBB domain to importin ␤ relieves the autoinhibition with the result that the affinity for NLS cargo increases into the nM range (24 -26).
In this paper, we have studied the molecular basis for the recognition of phospholipid scramblase 1 (PLSCR1) NLS by the adaptor importin ␣. PLSCR1 is a lipid-binding protein that was proposed to accelerate transbilayer movement of phospholipids at high calcium concentrations (27)(28)(29). In the cytoplasm PLSCR1 is localized mostly at the endofacial plasma membrane via multiple palmitoylations, where it binds calcium via a direct EF hand-like calcium binding domain (29). It was reported recently that PLSCR1 has the potential to enter the nucleus (30). This nuclear activity was observed either in the presence of inhibitors of palmitoyl-CoA, which block PLSCR1 anchoring to the plasma membrane, or by treatment with interferon ␣ (31), which is known to activate the PLSCR1 gene expression transcriptionally (32,33). In the nucleus PLSCR1 might function in the overall cellular response of those cytokines and growth factors that can induce expression of the PLSCR1 gene. The PLSCR1 NLS was mapped to the C terminus region 257 GKISKHWTGI 266 , and it was reported to be sufficient for nuclear import when conjugated to bovine serum albumin via association with the importin ␣/␤ heterodimer (30).
We have used x-ray crystallography to define the interaction between the import adaptor importin ␣ and the nonclassical PLSCR1 NLS. Remarkably, our data indicate that the PLSCR1 NLS binds the N-terminal region of importin ␣ between Arms 1 and 4. This region spans the classical NLS binding domain (Arm 3-4) and also includes a novel binding interaction with Arm 1-2.

Expression and Purification of Recombinant Proteins-
The gene encoding truncated importin ␣-2 (residues 70 -529) lacking the IBB domain (⌬IBB-importin ␣) was cloned in a pET-30a vector (Novagen) and expressed in Escherichia coli BL21 (DE3) strain. Purification was carried out as described previously (19). The SV40 NLS was cloned inframe to the green fluorescent protein (GFP) in a pET-28a vector. The fusion protein His 6 -SPKKKRKVEAS-NLS-GFP was expressed and purified as described previously (34). The PLSCR1 NLS fused to GFP (His 6 -SGKISKHWTGIEAS-GFP) was generated by ligating two oligonucleotides encoding the PLSCR1 NLS sequence into a unique Not site of the plasmid pET-28a-GFP. His 6 -SGKISKHWTGIEAS-GFP was expressed in E. coli BL21 (DE3) strain for 3 h at 30°C after induction with 1.0 mM isopropyl-␤-D-thiogalactopyranoside. Recombinant His 6 -SGKISKHWTGIEAS-GFP was purified by metal chelate affinity chromatography using nickel-nitrilotriacetic acid-agarose beads (Qiagen).
Fluorescence Anisotropy Assay-Fluorescence anisotropy measurements were carried out using a Spex FluoroMax-3 fluorometer (Horiba Group). Binding of SV40 NLS-GFP and PLSCR1 NLS-GFP to importin ␣ were measured in phosphate-buffered saline (135 mM NaCl, 2.7 mM KCl, 40 mM NaH 2 PO, 44 mM Na 2 HPO, pH 7.4) as described previously (34). Changes in GFP anisotropy as a function of the importin ␣ concentration were used to calculate the fraction of the GFP fluorophore bound, yielding a binding isotherm for the reaction. The binding isotherm was then fit through nonlinear regression and K d values calculated using the program Prism.
Crystallization of ⌬IBB-Importin ␣ Bound to the PLSCR1 NLS-For crystallization, pure ⌬IBB-importin ␣ was concentrated to ϳ22 mg/ml using a Millipore concentrator (molecular mass cutoff, 30 kDa). A peptide encompassing residues 257 GKISKHWTGI 266 of PLSCR1 NLS was synthesized by solid phase methods and purified by reverse phase high pressure liquid chromatography (molecular mass, 3,856.83 Da). Crystals of importin ␣-2 (73-529)-PLSCR1 NLS complex were grown using hanging drop vapor diffusion technique. Typically 2.3 l of ⌬IBB-importin ␣ was mixed with 0.7 l of PLSCR1 NLS peptide, and 3 l of reservoir solution containing 0.6 -0.7 M sodium citrate, 100 mM Hepes, pH 6.0, and 10 mM ␤-mercaptoethanol. Crystals grew at 20°C for 3 days. 25% glycerol was added to the crystals as cryoprotectant before flash-freezing in a nitrogen stream at Ϫ170°C.
Crystallographic Analysis-X-ray data were collected on beam-line 9 -2 at the SSRL on a Quantum-3 ϫ 3 CCD detector. Data were processed with the HKL suite (35) and analyzed further using CCP4 programs (36). Crystals of ⌬IBB-importin ␣-PLSCR1 NLS complex belong to space group P2 1 2 1 2 1 , with one complex/asymmetric unit. The structure was determined by molecular replacement using crystallography NMR software (CNS; 37) and importin ␣-2 (70 -529) as search model (PDB code 1EJL). Initial rigid body refinement of the model using program CNS (37) yielded working and free R factors of 30.49% and 29.93% (calculated using 10% of the observed reflections). The PLSCR1 NLS was built in F o Ϫ F c and 3F o Ϫ 2F c electron density difference maps with several rounds of manual building in O (38) followed by simulated annealing and grouped B factor with CNS (37). The final model, which includes all residues for importin ␣-2 (70 -529), PLSCR1 NLS (257-266), and 134 water molecules, has a R work of 22.9% and R free of 25.8%. The average B factors for importin ␣-2 (70 -529) and PLSCR1 NLS (257-266) are 28.77 and 48.043 Å Ϫ2 , respectively. Stereochemistry was checked by PROCHECK (36) and revealed no residues in disallowed regions. Data collection and refinement statistics are summarized in Table I. Structural figures were made using Bobscript (39) and Raster3d (40). Coordinates for the ⌬IBB-importin ␣-PLSCR1 NLS complex were deposited in the Protein Data Bank with accession code 1Y2A.

A Poorly Basic NLS Enriched in Hydrophobic
Residues-Ben-Efraim et al. (30) showed that PLSCR1 is imported into the nucleus by the importin ␣/␤ heterodimer (30) and the NLS sequence responsible for nuclear import comprises residues 257 GKISKHWTGI 266 . This sequence is less basic than classical SV40 NLS (the calculated isoelectric point is ϳ9.8 versus 11.38 for the SV40 NLS) and significantly more hydrophobic. Using the Kyte-Doolittle algorithm (41), we predicted that the PLSCR1 NLS is about three times more hydrophobic than the SV40 NLS (1.90 versus 0.62), mostly because of three bulky hydrophobic residues at positions Ile 259 , Trp 263 , and Ile 266 .
There are four distinct human PLSCR genes (hPLSCR1-4) as well as corresponding orthologous genes in mouse (mPLSCR1-4), and putative orthologs in Drosophila, Caenorhabditis elegans, and Saccharomyces cerevisiae (42). To determine whether the paucity of basic residues is a specific feature of the PLSCR1 NLS or if it is conserved in all phospholipid scramblases, we have aligned the primary sequence of the four human homologs (42). Despite significant differences in the length and sequence of these proteins, the NLS region is highly conserved (Fig. 1A). Four of the 10 residues are invariant in humans, although none of them is basic. These residues include two Gly 257 and Gly 265 as well as the two bulky hydrophobic amino acids Ile 259 and Trp 263 . Notably, only two basic amino acids, Lys 258 and Lys 261 , are conserved in three of the four human homologs, but not in PLSCR4 (Fig. 1A). This pattern of conservation is significantly different from that observed in importin ␣-dependent NLSs characterized to date (43). Consistently, phospholipid scramblases may have developed distinct mechanisms of binding to the importin ␣ which differ from those used by classical import cargoes.
PLSCR1 NLS Binds Importin ␣ with High Affinity-To characterize the recognition of the PLSCR1 NLS by importin ␣, we first determined the binding affinity of the minimum PLSCR1 NLS 257 GKISKHWTGI 266 for the import adaptor importin ␣. The binding study was carried out using the fluorescence depolarization anisotropy assay developed by Hodel et al. (34) to study the binding of classical NLSs to importin ␣. Briefly, the PLSCR1 NLS region 257 GKISKHWTGI 266 was fused to the GFP, and the fusion protein PLSCR1 NLS-GFP was assayed for binding to a truncated version of importin ␣-2 (residues 70 -529) which lacks the IBB domain. The increase of GFP anisotropy that results from the binding of PLSCR1 NLS to ⌬IBB-importin ␣ was measured as a function of ⌬IBB-importin ␣ concentration, yielding a K d ϳ45.8 nM (Fig. 1B). This value is comparable with that observed for the binding of full-length PLSCR1 to importin ␣ (30), thus confirming that region 257-266 genuinely represents the minimum NLS of PLSCR1. As a positive control, we used the SV40 NLS-GFP, which, under identical experimental conditions was found to bind ⌬IBBimportin ␣ with K d ϳ10 nM (Fig. 1B). This value agrees well with the dissociation constants measured using Biacore or microtiter binding assay (43), suggesting that the apparent affinities derived by our fluorescence anisotropy-based binding assay are consistent with those estimated using other methods.
Crystallographic Analysis of Importin ␣ Bound to the PLSCR1 NLS-We have used x-ray crystallography to determine the molecular basis of the interaction between PLSCR1 NLS and importin ␣. As the IBB domain is not required for this interaction, we have crystallized a truncated construct of importin ␣ (residues 70 -529) in complex with a chemically synthesized peptide comprising the PLSCR1 NLS sequence 257 GKISKHWTGI 266 . Crystals of the complex were grown in the presence of a 2-fold molar excess of PLSCR1 NLS, using 0.6 M sodium citrate at pH 6.0. Using synchrotron radiations, we found that ⌬IBB-importin ␣-PLSCR1 NLS complex crystals diffract x-ray beyond 2 Å resolution ( Table I). The structure was determined by molecular replacement, using mouse ⌬IBBimportin ␣-2 (70 -529) (PDB code 1EJL) as a search model. The PLSCR1 NLS was built in unbiased F o Ϫ F c and 3F o Ϫ 2F c electron density difference maps calculated using phase angles from the refined importin ␣ model and structure factor amplitudes derived from the observed x-ray data. After several rounds of manual building and refinement all PLSCR1 NLS residues were clearly modeled in the electron density. The final model, which includes residues (70 -529) of importin ␣, (257-266) of PLSCR1 NLS, and 134 water molecules, was refined to a working and free R factors of 22.9 and 25.8%, respectively, using all diffraction data between 40 and 2.2 Å. A ribbon diagram of the ⌬IBB-importin ␣-PLSCR1 NLS complex structure is shown in Fig. 2A. The Arm core of importin ␣ consists of nine Arm repeats that pack into a right-handed superhelix of helices. As observed previously for yeast and vertebrate importin ␣ (18,20,22), Arm 1-2 and 9 -8 are slightly curved in opposite directions with respect to a central helical core formed by Arm repeats 3-7. The nonclassical PLSCR1 NLS is bound to the N terminus of importin ␣ where it interacts with Arm repeats 1-4. In contrast to previous complexes of importin ␣ with monopartite NLSs (18 -20), a single copy of the PLSCR1 NLS is bound to importin ␣ between Arms 1 and 4, in a region that includes the major NLS binding sites located between Arms 2 and 4. No significant density is seen at the minor binding site (Arm 2-4). The final PLSCR1 NLS model (Fig. 2B) has an average B factor of 48Å Ϫ2 , slightly higher than importin ␣ (28.8 Å Ϫ2 ). The former value is in line with the B factor observed for the classical SV40 NLS, which was determined to a comparable resolution of 2.2 Å in complex with yeast importin ␣ (20). The conformation of importin ␣ adopted upon binding to the PLSCR1 NLS is virtually identical to those in complex with both classical monopartite NLS (18 -21) and bipartite NLS (19,20). This suggests that the binding of the PLSCR1 NLS is not accompanied by significant conformational changes in the protein.
The binding interface between importin ␣ and the PLSCR1 NLS is remarkably extended. The NLS buries ϳ654.4 Å 2 of surface between Arm repeats 1 and 4. The PLSCR1 peptide is bound in a fully extended conformation, which resembles the binding of the nonclassical parathyroid hormone-related protein NLS to the import receptor importin ␣ (14). We identify two distinct areas of contact between importin ␣-2 and the PLSCR1 NLS (Fig. 2C). First, the N-terminal moiety 257 GKISK 262 occupies the conventional NLS binding site between Arm 2 and 4 which is transversed to the importin ␣ superhelical axis. In analogy with the nomenclature used to describe classical SV40 NLS residues, we will refer to these residues as P1-P5. Second, the C-terminal sequence PLSCR1 262 HWTGI 266 spans Arm repeats 1-2 where it binds importin ␣ laterally, adopting a conformation that is almost parallel to the superhelical axis. These residues will be indicated as S1-S5 (Fig. 3A).
Interaction of Importin ␣ with the N-terminal PLSCR1 Sequence 257 GKISK 262 -The N-terminal PLSCR1 NLS moiety 257 GKISK 262 spans the major NLS binding site between Arm repeats 2 and 4 (19,22). As seen in the recognition of the SV40 NLS (18,19), the NLS backbone is held in the binding groove by intimate contacts with three conserved asparagines Asn 235 , Asn 188 , and Asn 145 . These residues make hydrogen bonding interactions with the main chain amide groups of PLSCR1 The difference map (contoured at 1.0 above noise) was computed using refined phase angles from the importin ␣ model and observed structure factor amplitudes calculated for the x-ray data. C, schematic diagram of the interactions between importin ␣ and the PLSCR1 NLS. Black and purple dashed lines indicate binding between importin ␣ side chains and PLSCR1 NLS side chain and main chain, respectively. Intra-and intermolecular contacts between atoms in distance range 2.5-4.5 Å were determined using the program CNS (37). residues at position P1, P3, and P5, respectively (Fig. 2C). The PLSCR1 residues occupying positions P1-P5 are dramatically different from the SV40 NLS polybasic stretch 127 KKKRK 131 . At position P1 an invariant residue Gly 257 is probably important to provide high conformational flexibility to the following Lys 258 , which, together with Lys 261 , occupy the conventional positions annotated as P2 and P5, respectively (Fig. 2C). Both PLSCR1 Lys 258 and Lys 261 contact importin ␣ intimately by inserting their N-terminal basic group inside acidic pockets of the protein. Whereas the basic N-terminal group of Lys 258 directly contacts the C-terminal group of Asp 192 , Lys 261 at position P5 makes a salt bridge with ␣Glu 181 (Fig. 3A). Notably, three tryptophans surround the PLSCR1 Lys 261 , two from importin ␣, Trp 184 and Trp 142 , and one, Trp 263 , from the PLSCR1 (Fig. 3B). This tryptophan "cage" locks the ⑀-N-terminal group of Lys 261 in a conformation ideal to interact with ␣Glu 181 . However, there is a significant difference in the chemical interactions engaged by the three tryptophans. Whereas the indole rings of ␣Trp 184 and ␣Trp 142 point directly toward the ⑀-N-terminal groups of PLSCR1 Lys 261 , PLSCR1 Trp 263 is oriented more toward the aliphatic side chain of Lys 261 . The former interaction is consistent with a cation-interaction, whereas the latter is presumably more a hydrophobic contact. To investigate further this prediction, we used the program CaPTURE (Cation-Trends Using Realistic Electrostatics) (44), which predicts and quantifies cation-interactions based on ab initio calculations from three-dimensional structures. Interestingly, CaPTURE predicts that both ␣Trp 184 and ␣Trp 142 engage in energetically significant intermolecular cat-ion-interactions with Lys 261 . The energetic contributions were estimated to be approximately Ϫ7.40 and Ϫ2.41 kcal/mol for ␣Trp 184 and ␣Trp 142 , respectively. As predicted by visual examination of the three-dimensional structure, the potential energetic contribution resulting from cation-interaction for the pair Trp 263 -Lys 261 was not significant (Ͻ2 kcal/mol), as the ring of the tryptophan points more toward the aliphatic portion of the side chain of the Lys 261 than to the cationic N-terminal group.
Finally, PLSCR1 residues Ile 259 and Ser 260 at position P3 and P4 show a pattern of molecular interactions completely different from classical SV40-like sequences. Instead of basic Arg/Lys, P3 presents the aliphatic side chain of residue Ile 259 , which makes strong hydrophobic contacts with the indole rings of ␣Trp 231 and ␣Trp 184 . In turn, the hydroxyl group of Ser 260 at position P4 makes hydrogen bond to importin ␣Ser 149 (Fig. 2C).
Interaction of Importin ␣ with the C-terminal PLSCR1 Sequence 262 HWTGI 266 -The NLS region upstream the P5 site includes the sequence 262 HWTGI 266 . Here three of five residues (Trp 263 , Gly 265 , and Ile 266 ) are both highly conserved and nonbasic, strengthening the idea that this region represents a strong nonbasic determinant for nuclear import (Fig. 1A). Similar to the C-terminal moiety, the molecular recognition of this heptapeptide is based on both main chain and side chain contacts with importin ␣. The 263 WTGI 266 main chain is held in place by four contacts with importin ␣ residues, Gln 98 , Arg 101 , Lys 102 , and Ser 105 (Fig. 2C). These residues orient the PLSCR1 backbone along the binding site, in a way analogous to the asparagine arrays observed in the major NLS binding site. Intriguingly, analysis of the importin ␣ primary sequence conservation indicates that all four residues, Gln 98 , Arg 101 , Lys 102 , and Ser 105 , are highly conserved within vertebrates. This suggests that the extended binding interaction observed for the PLSCR1 NLS may not be a unique feature of this cargo and that Arm 1-2 potentially represent a general recognition site for nonclassical NLSs.
In addition to backbone contacts, importin ␣ makes at least three specific side chain-side chain contacts with PLSCR1 NLS residues (Fig. 2C). Notably, PLSCR1 His 262 and Thr 264 at positions S1 and S3 make electrostatic interactions with ␣Glu 107 and ␣Arg 101 , respectively. In turn, the conserved Trp 263 at S2 engages in an intramolecular interaction with the aliphatic portion of Lys 261 (Fig. 3B), which, as described previously, sits at position P5.
In Vivo Analysis of the PLSCR1 NLS-Ben-Efraim et al. (30) have shown that the classical SV40 NLS competes with PLSCR1 NLS in a nuclear import assay in digitonin-permeabilized cells, suggesting that PLSCR1 NLS might bind to importin ␣ in the same fashion as does the SV40 NLS (30). This hypothesis is fully supported by our three-dimensional structure. The structural alignment in Fig. 4 shows that the conformation adopted by PLSCR1 NLS main chain between residues 257 and 261 ( 257 GKISK 261 ) is virtually identical to that observed in classical NLSs (Fig. 4). In addition, Lys 258 and Lys 261 align well with the canonical P2 and P5 sites of other NLS peptides solved in complex with importin ␣ (18 -21) (Table II). However, as discussed previously, PLSCR1 NLS residues at position P1, P3, and P4 are nonbasic, as well as the N-terminal sequence 262 HWTGI 266 neither matches nor aligns to any NLS sequence discovered to date.
To characterize further the PLSCR1 NLS in vivo, we have studied the effect of single point mutations in the PLSCR1 NLS in a transient transfection assay. SVT2 fibroblasts were transfected with cDNA for wild-type PLSCR1 or the PLSCR1 ( 184 AAAPAA 189 ) mutant, which bears mutations of the palmitoylation site, preventing trafficking of this protein to the plasma membrane (31). Expressed PLSCR1 was detected with the murine monoclonal antibody 4D2 specific for human PLSCR1 (32,33). As reported previously, wild type PLSCR1 showed a distinctive immunofluorescent staining at the endofacial surface of the plasma membrane, where the protein is anchored via multiple palmitoylations (Fig. 5, top panel). To characterize the importance of single conserved residues, we repeated the transfection assay using the nuclear trafficking PLSCR1 ( 184 AAAPAA 189 ) mutants with additional alanine substitutions at residues Lys 258 , Lys 261 , and Trp 263 . Notably, none of these single point mutations was found to abolish nuclear localization of PLSCR1 fully (Fig. 5). Although single point mutations in the PLSCR1 NLS did not significantly reduce its nuclear accumulation, the dominant negative mutant ( 267 GAISAAWTGI 288 ) (30) was found to be completely excluded from the cell nucleus. Taken together these transfection data indicate that the PLSCR1 NLS is not disrupted by a single point mutation at position P2 (or P5). This is in contrast to classical SV40-like NLS sequences, where a single point mutation at position P2 can reduce the binding affinity by 100-fold (34), which results in a complete loss of nuclear import (45). DISCUSSION The prototypical NLS of the SV40 large T antigen consists of a single cluster of five contiguous positively charged residues ( 126 PKKKRKV 132 ) which is recognized by transport adaptor importin ␣. Although other nuclear localization signals such as the human c-myc NLS ( 320 PAAKRVKLD 328 ) are somewhat less basic, a clustering of consecutive three to five positive charged residues appears to be the main diagnostic feature of most nuclear import sequences. Moreover, a conserved lysine at position P2 (Lys 128 in the classical SV40 NLS) is crucial for nuclear import. Mutation of this residue completely abolishes nuclear import (45,46) and reduces the binding affinity for importin ␣ by more than 100-fold (34). Mutation of one of the adjacent basic residues at P1, P3, P4, or P5 reduces but does not completely abolish nuclear import (46).
We have studied the molecular basis for the recognition of poorly basic PLSCR1 NLS by importin ␣. This poorly basic NLS deviates significantly from the polybasic SV40-like NLS consensus, but binds importin ␣ with high affinity (K d ϳ45 nM). Our structural analysis indicates that despite the paucity of basic amino acids, PLSCR1 NLS recognition of importin ␣ is intimate and extensive. A complex network of side chain and main chain interactions is seen in the crystal structure, which spans Arm repeats 1-4 of the adaptor. Multiple structural determinants contribute to the recognition of the nonclassical PLSCR1 NLS. First, the entire N-terminal portion of the importin ␣ Arm core domain provides an extended binding surface that extends and orients the PLSCR1 NLS backbone along the classical NLS binding site (Arm 2-4) as well as on the lateral surface of Arm 1-2. The PLSCR1 NLS backbone is contacted by a complex set of hydrogen bonding with importin ␣ side chains. These interactions are largely mediated by three conserved asparagines within Arm 2-4 and four conserved importin ␣ residues, Gln 98 , Arg 101 , Lys 102 , and Ser 105 , between Arm 1-2. Second, two basic residues at position P2 and P5 provide the essential electrostatic binding complementarily with importin ␣. Notably, both Lys 258 and Lys 261 are flanked by two small amino acid at positions P1 (Gly 257 ) and P4 (Ser 260 ). This may reduce the steric hindrance introduced by a bulky side chain and contribute to provide higher conformational flexibility to basic side chains of Lys 258 and Lys 261 , hence enhancing their binding specificity for importin ␣. Although the recognition of Lys 258 at positions P2 closely resembles that described previously in the recognition of the SV40 NLS (13)(14)(15), our in vivo assay demonstrates that this residue is important but not essential for nuclear import. Third, hydrophobic residues in the PLSCR1 NLS engage in intramolecular contacts within the NLS as well as intramolecular contacts with importin ␣. For instance, the hydrophobic side chain of PLSCR1 Ile 259 is held in place by the two ␣Trp 184 and ␣Trp 231 . Similarly, PLSCR1 Trp 263 packs in close proximity (ϳ4 Å) to the aliphatic portion of Lys 261 , sealing the lysine side chain in a trypthophan cage.
The structural prediction that multiple binding determinants provide high affinity binding specificity to importin ␣ in the absence of a cluster of positively charged residues is also corroborated by our in vivo transfection assay. As mentioned previously, a well accepted paradigm in the nucleocytoplasmic transport field lies in the absolute necessity of a Lys at position P2 (45,46). A large body of work has shown that this residue is responsible for most of the binding energy (34). Deletion or substitution of this residue dramatically reduces the binding affinity. In contrast, our in vivo mutational data reveal that a single point mutation of P2 or P5 or at the conserved Trp 263 can reduce nuclear import of PLSCR1 but not abolish it. At least three mutations are necessary to render the PLSCR1 distribution fully cytoplasmic. The somewhat dispensable role of Lys 258 at position P2 is also indirectly confirmed by the observation that human PLSCR4 lacks a basic residue at this position (Fig.  1A). The only two basic side chains for this phospholipid scramblase homolog are found at position P5 and S1. However, it should be pointed out that the PLSCR4 NLS remains putative, as the nuclear import of PLSCR4 has not been demonstrated.
The ability of the importin ␣ binding groove to orient the PLSCR1 NLS backbone and efficiently contact its side chains suggests that the clustering of positively charged residues within classical NLS sequence is not a mandatory feature for all nuclear import signals. For instance, both PLSCR1 residues at positions P3 and P4 are nonbasic, yet still efficiently bound to importin ␣ via a combination of hydrophobic and polar contacts. This emphasizes the idea that the recognition of distinct NLS sequences is highly plastic. The Arm scaffold represents a versatile binding surface that can efficiently accommodate diverse sequences. By fixing and orienting the NLS backbone, importin ␣ can recognize different NLS sequences and maximize the number of side chain contacts. As seen in the recognition of the PLSCR1 NLS, the moderately basic sequence 257 GKISK 262 is bound within the major NLS binding site, whereas the nonbasic hydrophobic region 262 HWTGI 266 is accommodated on the lateral surface of Arm 1-2. Such backbonecontrolled binding maximizes the number of side chain contacts between the protein and the NLS, thus achieving high binding specificity. Energy Sciences, for assistance with synchrotron data collection. The SSRL Structural Molecular Biology Program is supported by the Department of Energy, Office of Biological and Environmental Research, and by the National Institutes of Health, National Center for Research Resources, Biomedical Technology Program, and the National Institute of General Medical Sciences. The technical assistance of Lilin Li, Jo-Lawrence Bigcas, and Daniela Junquiera is gratefully acknowledged as are the advice, suggestions, and assistance of Drs. Ji Zhao and Quansheng Zhou.