Molecular Characterization of the Ran-binding Zinc Finger Domain of Nup153*

The nuclear pore complex is the gateway for selective traffic between the nucleus and cytoplasm. To learn how building blocks of the pore can create specific docking sites for transport receptors and regulatory factors, we have studied a zinc finger module present in multiple copies within the nuclear pores of higher eukaryotes. All four zinc fingers of human Nup153 were found to bind the small GTPase Ran with dissociation constants ranging between 5 and 40 μm. In addition a fragment of Nup153 encompassing the four tandem zinc fingers was found to bind Ran with similar affinity. NMR structural studies revealed that a representative Nup153 zinc finger adopts the same zinc ribbon structure as the previously characterized Npl4 NZF module. Ran binding was mediated by a three-amino acid motif (Leu13/Val14/Asn25) located within the two zinc coordination loops. Nup153 ZnFs bound GDP and GTP forms of Ran with similar affinities, indicating that this interaction is not influenced by a nucleotide-dependent conformational switch. Taken together, these studies elucidate the Ran-binding interface on Nup153 and, more broadly, provide insight into the versatility of this zinc finger binding module.

The nuclear pore complex is the gateway for selective traffic between the nucleus and cytoplasm. To learn how building blocks of the pore can create specific docking sites for transport receptors and regulatory factors, we have studied a zinc finger module present in multiple copies within the nuclear pores of higher eukaryotes. All four zinc fingers of human Nup153 were found to bind the small GTPase Ran with dissociation constants ranging between 5 and 40 M. In addition a fragment of Nup153 encompassing the four tandem zinc fingers was found to bind Ran with similar affinity. NMR structural studies revealed that a representative Nup153 zinc finger adopts the same zinc ribbon structure as the previously characterized Npl4 NZF module. Ran binding was mediated by a three-amino acid motif (Leu 13 /Val 14 / Asn 25 ) located within the two zinc coordination loops. Nup153 ZnFs bound GDP and GTP forms of Ran with similar affinities, indicating that this interaction is not influenced by a nucleotidedependent conformational switch. Taken together, these studies elucidate the Ran-binding interface on Nup153 and, more broadly, provide insight into the versatility of this zinc finger binding module.
In eukaryotic cells, nucleocytoplasmic transport is a critical function for maintaining basic processes such as cell cycle regulation and gene expression. Transport across the nuclear envelope occurs through nuclear pore complexes, large macromolecular structures of ϳ125 MDa that are embedded in the nuclear envelope. Nuclear pore complexes are comprised of three major architectural features: 1) a central framework, which includes integral membrane proteins and resides within the plane of the membrane; 2) a cytoplasmic ring and its extended filaments; and 3) a nuclear basket, which is composed of distal and proximal rings connected by eight fibers (reviewed in Ref. [1][2][3]. Proteomic approaches have revealed that, relative to its large size, the nuclear pore complex is composed of only a small number of nucleoporins (Nups) 4 that are present in at least 8, and often 16 or 32, copies at the pore (4,5). More recently, bioinformatics and structural analyses have further indicated that a limited repertoire of structural motifs are found within the 30 proteins that form the macromolecular pore complex (3,6,7).
One distinctive nucleoporin domain that has not been examined structurally is a zinc finger domain present in multiple copies in both Nup153 and Nup358/RanBP2 (referred to herein as Nup358). These two proteins localize to different sites on the pore, with Nup358 situated on the cytoplasmic filaments and Nup153 positioned on the nuclear basket. Each protein functions in nucleo-cytoplasmic transport (Refs. 8 -16 and reviewed in Ref. 17) and in breakdown of the nuclear envelope at mitosis (18,19). Nup358 takes on additional roles at the kinetochore during mitosis (20,21).
The Nup153 and Nup358 zinc fingers are the defining members of the "RanBP2-type" zinc finger family (22). RanBP2-type zinc fingers conform to the consensus sequence pattern: W-X-C-X(2,4)-C-X(3)-N-X(6)-C-X (2)-C. This module is found in a functionally diverse population of proteins, with more than 1055 matches in 727 polypeptides using the ScanProsite tool. A similar consensus sequence that omits the initial tryptophan and allows only two residues between the first two cysteines defines a related polypeptide family termed Npl4 zinc fingers (NZF) (23,24). Three-dimensional structures of several different RanBP2/NZF fingers have revealed that the module forms two orthogonal ␤-hairpins that chelate a single zinc atom at one end of a small barrel (24 -27). The sequence conservation can be rationalized because each conserved residue performs an important structural role: the four cysteine residues coordinate zinc, the Trp residue forms the hydrophobic core of the module, and the Asn residue bridges strands 2 and 3. The small RanBP2/NZF module is a "stripped down" member of the much larger zinc ribbon family, whose members all utilize two ␤-hairpins to coordinate a single zinc ion between two Cys 2 "knuckles." Zinc ribbon proteins diverge significantly beyond the zinc coordination site, however, and frequently have complex structural elements inserted between the two knuckles (28).
Importantly, NZF modules can bind a wide variety of different macromolecules. For example, NZF modules in Npl4 (23), Vps36 (25), and Tab2/Tab3 (29) all bind ubiquitin (Ub) and allow these proteins to function in Ub-dependent pathways. A second NZF module in Vps36 binds Vps28 (30), bridging complexes involved in protein sorting, and the zinc finger in the transcriptional elongation factor TFIIS binds RNA (31). A better understanding of how different NZF family members recognize their cognate binding partners will be important for defining, and ultimately predicting, ligand specificity in this large protein family.
The zinc fingers found in Nup153 and Nup358 (hereafter referred to as Nup ZnF) are distinctive in that they are found in repetitive arrays. In addition to conforming to the RanBP2-type consensus, these zinc fingers exhibit additional sequence conservation at several non-structural positions, particularly Leu 13 (64% conserved), Val 14 (79%), and Ala 25 (79%) (numbering schemes are given in Figs. 1 and supplemental Fig. S1 and conservation is calculated for the representative sequences shown in Fig. S1). The equivalent three positions define the ligand binding site in the Npl4 zinc finger motif (25). Notably, Nup ZnF modules are present only in higher eukaryotes, suggesting a specialized role(s) in nuclear pore function. Consistent with this idea, the Nup ZnF regions of Nup153 and Nup358 are important in orchestrating mitotic nuclear envelope breakdown, a type of nuclear membrane remodeling unique to higher eukaryotes. In this context, the Nup ZnFs are proposed to serve as a scaffold for recruitment of the coatomer complex COPI (18,19). Other proteins that associate with Nup ZnF regions include the small GTPase Ran and the transport receptor exportin 1 (32)(33)(34).
Given the pivotal role that Ran plays in nucleocytoplasmic trafficking (reviewed in Refs. 3 and 35)), Ran binding by the Nup ZnF regions may be important in helping to regulate traffic through the higher eukaryotic pore. As a Ras-like GTPase, Ran fits the paradigm in which GTP and GDP binding create distinct conformational states, allowing Ran to serve as a molecular switch (36 -38) (reviewed in Ref. 39). Previous studies have suggested that RanGDP, but not RanGTP, binds specifically to the Nup ZnF regions (32,34), which is of potential interest because the roles of Ran often depend critically on its nucleotide state.
To begin to investigate how Nup ZnF modules function in nuclear pore biology, we have determined the three-dimensional structure of a nucleoporin zinc finger and characterized its Ran binding properties.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Plasmids used in this study are listed in supplemental Table SII. The expression construct for wildtype ZnF2 from human Nup153 was created by cloning annealed oligonucleotides 3 and 4 (Table SIII) into the NdeI/ BamHI sites of a pGEX2T vector modified to contain a TEV cleavage site between the GST and ZnF2 polypeptides (WISP01-69). Inserts for human Nup153 zinc fingers 1, 3, and 4 were created by PCR amplification from full-length Nup153 and subcloned into WISP01-69 using NdeI/XhoI. All site-directed mutations were created using QuikChange TM mutagenesis (Stratagene). For the Nup153 ZnF constructs L13T and V14F mutations were first introduced using oligonucleotides 9 and 10 (Table SIII), and this construct was then mutated further to alter Ala 25 to Met using oligonucleotides 11 and 12 (Table  SIII). Expression constructs for wild-type Npl4 NZF domain and the T13L,F14V mutant were described previously (24,25). The latter construct was mutated further to create the T14L,F15V,M25A mutant using oligonucleotides 13 and 14 (Table SIII). The GST-Ran construct was a kind gift from Sally Kornbluth (Duke University). Site-directed mutation of this construct to create GST-RanQ69L was performed using oligonucleotides 17 and 18 (Table SII). His-tagged importin ␤ fragments 1-462 and 45-462 in pQE60 (Qiagen) were a kind gift from Dirk Görlich (University of Heidelberg) (40).
Expression and Purification of Recombinant Ran-To produce Ran for use as the analyte in biosensor experiments, 10-liter cultures of BL21(DE3) RIL cells (Stratagene), transformed with Ran expression vectors, were grown at 37°C in a fermenter to an A 600 of ϳ0.8. Isopropyl 1-thio-␤-D-galactopyranoside was then added to 1 mM and the culture grown for an additional ϳ15 h at 22°C. Following centrifugation, pellets from 2 liters of culture were resuspended in 100 ml of Buffer A (25 mM Tris, pH 8.0, 500 mM NaCl, 5 mM MgCl 2 , 5 mM ␤-ME, 1 mM DTT). All purification steps were performed at 4°C. Bacteria were lysed by sonication and the lysate was centrifuged to remove insoluble debris. GST-Ran was recovered from the cleared lysate by fast protein liquid chromatography using a GSTPrep FF 16/10 affinity column (Amersham Biosciences). After washing with 4 column volumes of buffer A, protein was eluted with 20 mM glutathione in 25 mM Tris, pH 8.0, 200 mM NaCl, 5 mM MgCl 2 , 5 mM ␤-ME, and 1 mM DTT. Eluted protein was then dialyzed overnight into 25 mM Tris, pH 8.0, 50 mM NaCl, 5 mM MgCl 2 , 2.5 mM CaCl 2 , and 1 mM DTT in the presence of thrombin (Calbiochem; ϳ0.5 units/mg GST-Ran). Ran was purified away from GST by sequential anion and cation chromatography steps. Flow-through fractions from a DEAE-Sepharose column were applied to an S-Sepharose column. A salt gradient (25 mM Tris, pH 8.0, 5 mM MgCl 2 , 1 mM DTT, and 50 -1000 mM KCl) was used to elute protein from the S-Sepharose column, but Ran largely eluted in 50 mM KCl fractions. Material from these fractions was concentrated and residual GST and GST-Ran were removed by passage over 1.5 ml of glutathione-Sepharose. Purified Ran was loaded with nucleotide (see below), flash frozen in liquid nitrogen, and stored at Ϫ80°C. This procedure typically resulted in a yield of ϳ10 mg of Ran per 2-liter pellet.
Thrombin cleavage of GST-Ran resulted in 15 non-native N-terminal amino acid residues (NH 2-GSPGISGGGGGILDS). Precise cleavage by thrombin was confirmed by matrix-assisted laser desorption ionization mass spectrometry (calculated mass, 25,635 Da; observed mass, 25,634 Da). Thrombin-cleaved GST-Ran was used for the binding experiments reported herein. A GST-Ran construct with a TEV protease cleavage site was also cloned and purified ( Fig. S2 and supplementary methods) to produce a protein with only one non-native N-terminal amino acid residue (NH 2 -S). The Nup153 ZnF binding affinities of this shorter Ran construct were similar to those of the longer thrombin-cleaved Ran protein (data not shown). Purities and concentrations of Ran preparations were analyzed by SDS-PAGE and amino acid analyses, respectively.
GST-Ran immobilized on biosensor chips was prepared slightly differently: pellets from 1-liter cultures were resuspended in 40 ml of Buffer B (10 mM NaH 2 PO 4 , pH 7.4, 150 mM NaCl, 5 mM MgCl 2 , 5 mM ␤-ME with 1/2 tablet of Complete protease inhibitor (Roche Applied Science)) and sonicated. Cleared lysates were applied to a 1.5-ml bed volume of glutathione-Sepharose (Amersham Biosciences). The column was washed with 10 bed volumes of Buffer B and eluted with 20 mM glutathione in 25 mM Tris, pH 7.0, 5 mM MgCl 2 , 2 mM DTT, and 10 M ZnCl 2 . Eluates were dialyzed overnight against the same buffer without glutathione and then concentrated to 1-3 ml prior to loading with nucleotide and biosensor analysis.
Expression and Purification of Recombinant Zinc Finger Domains-To prepare pure Nup153 ZnF for biosensor binding experiments, BL21(DE3) RIL cells containing the Nup153 ZnF2 expression construct were grown at 37°C in a fermentor to an A 600 of ϳ0.5. Protein expression was induced with isopropyl 1-thio-␤-D-galactopyranoside (0.5 mM) and protein allowed to accumulate for an additional 4 h. Following centrifugation, a pellet from 6 liters of culture was resuspended in ZnF Buffer (10 mM NaH 2 PO 4 , pH 7.4, 150 mM NaCl, 5 mM ␤-ME, 10 M ZnCl 2 ). All purification steps were performed at 4°C. Bacteria were lysed by sonication and the lysate was centrifuged (40,000 ϫ g for 30 min) to remove insoluble debris. GST-Nup153 ZnF2 was recovered from the cleared lysate by affinity chromatography using a GSTPrep FF 16/10 affinity column (Amersham Biosciences). Bound protein was washed with 10 column volumes of ZnF buffer and eluted with 20 mM glutathione in 50 mM Tris, pH 8.0, 5 mM ␤-ME, and 10 M ZnCl 2 . Protein fractions were pooled and dialyzed for 16 h into 50 mM Tris, pH 8.0, 100 mM NaCl, 10 M ZnCl 2 , 1 mM DTT in the presence of recombinant TEV protease (10 g/mg GST-Nup153 ZnF2). The protein solution was concentrated to ϳ3 ml, and residual GST-Nup153 ZnF2 and GST were removed by Superdex-75 gel filtration chromatography (Amersham Biosciences) in 20 mM Tris, pH 7.5, 150 mM NaCl, 5 mM ␤-ME, 2 mM DTT, and 10 M ZnCl 2 . TEV protease cleavage left three non-native residues at the N terminus of ZnF2 (NH 2 -GHM), but the first native amino acid was designated Val 3 to keep the numbering consistent with the NZF consensus (24).
GST-Nup153 ZnF fusion proteins were captured onto biosensor chips directly from fresh, cleared bacterial lysates. Protein expression was induced with 1 mM isopropyl 1-thio-␤-Dgalactopyranoside (0.8 -1 A 600 , 5 h, 25°C). Pellets from 2 ml of culture were resuspended in either B-PER lysis buffer (Pierce) or Buffer B supplemented with 10 M ZnCl 2 (most experiments used the latter). Cells resuspended in B-PER buffer were rotated at room temperature for 10 min, followed by sonication and centrifugation. Cells resuspended in Buffer B were supplemented with lysozyme (1 mg/ml) and incubated on ice for 20 min, followed by sonication and centrifugation.
Nucleotide Loading of Ran-Nucleotide (2 mM) and EDTA (10 mM) were added to Ran preparations and the samples were rotated at room temperature for 30 min. MgCl 2 was then added to a final concentration of 20 mM (41).
Preparation of Samples for NMR Spectroscopy-Unlabeled human ubiquitin and 15 N-and 15 N/ 13 C-labeled Nup153 ZnF2 were expressed and purified as described previously (24) with the exception that the TEV protease cleavage buffer used included 5 mM sodium citrate and 5 mM ␤-ME in place of DTT. Purifications typically yielded 4 -5 mg of pure Nup153 ZnF2 and 40 mg of ubiquitin per liter of culture. Samples for structure determinations were 1 mM labeled protein in NMR buffer (20 mM D 11 -Tris, 50 mM NaCl, 2 mM D 6 -␤ME, 10 M ZnCl 2 , pH 7.0, in 90% H 2 O, 10% 2 H 2 O). Samples were degassed and flamesealed under argon in NMR sample tubes to reduce cysteine oxidation.
NMR Data Collection and Resonance Assignments-All NMR spectra were collected at 20°C on a Varian Inova 600 MHz spectrometer equipped with a triple-resonance 1 H/ 13 C/ 15 N Cold-Probe and z-axis pulsed field gradient capability. Backbone resonances were assigned using a suite of triple resonance experiments as described previously (24), employing two-dimensional versions of the HNCACB (42), CBCACONH, HNCO, and HN(CA)CO (43). Side chain assignments were completed using three-dimensional H(CCO)NH-TOCSY, (H)C(CCO)NH-TOCSY (44), HCCH-TOCSY, HCCH-COSY, and 15 N-edited TOCSY experiments. Aromatic resonances were assigned using a combination of 1 H, 13 C-HSQC, 13 C-edited NOESY experiments centered on the aromatic carbon resonances (125 ppm), a 13 C-edited NOESY experiment centered on the aliphatic region, and heteronuclear correlation experiments that correlate the C␤ carbons to the C␦ and C⑀ protons of the aromatic rings (45). Stereospecific assignments for ␤-methylene protons and 1 dihedral angle estimates were obtained using a combination of HNHB, HN(CO)HB (46), 15 Nedited TOCSY, and NOESY data (47). Stereospecific assignments of side chain methyl groups and qualitative determination of 1 and 2 dihedral propensities were obtained using long range carbon-carbon and carbon-proton couplings observed in LRCC (48) and LRCH (49) experiments. Three-dimensional 15 N-edited NOESY-HSQC (50, 51) and 13 C-edited NOESY-HSQC (150 ms mixing times) were used to generate distance restraints for refinement. Three-bond coupling constants ( 3 J HN-HA ) were obtained from a three-dimensional HNHA experiment (52) and / dihedral restraints were derived from backbone H␣, C␣, CO, C␤ chemical shifts using TALOS (53).
Structure Determination-Backbone and side chain correlations were assigned and NOE intensities were integrated using tools in SPARKY. 5 The solution structure of Nup153 ZnF2 was refined using automated NOE assignments and torsion angle dynamics as implemented in CYANA (version 2.1) (54,55). Initial refinements used NOE data alone to define the overall fold. Once the zinc coordination geometry was established, the final refinements added restraints for zinc-sulfur distances, cysteine-S␥ to S␥ distances, Zn to C␤ distances (14 total), dihedral restraints (25), hydrogen bonds (12), and stereospecific assignments (20). A total of 100 randomized conformers were "folded" into three-dimensional structures by including NOE constraints iteratively using criteria defined by CYANA. The 20 structures with the lowest CYANA target function were chosen for analysis. Structures were validated using PROCHECK-NMR (56), Aqua (56), and the programs supplied at the PDB deposition site. All figures were created using PYMOL (70).
NMR Chemical Shift Mapping of the Nup153 ZnF2:Ran Interface-The RanGDP binding interface on Nup153 ZnF2 was mapped by comparing the chemical shifts in independent NMR samples of 15 (57,58).
Biosensor Binding Studies-Binding studies were performed at 20°C using optical binding sensors (Biacore 2000 or Biacore 3000; Biacore AB, Uppsala, Sweden) equipped with researchgrade CM4 or CM5 sensor chips. Antibodies directed against GST were immobilized on all four flow cells using amine-coupling chemistry (2800 -6900 response units; response units reflect the change in refractive index due to the protein mass associated with the chip) (59). Cleared lysates of Escherichia coli expressing GST-ZnF proteins or purified GST-Ran proteins were captured individually on three anti-GST surfaces (640 -3280 response units) and recombinant GST was captured in parallel on the fourth antibody surface (460 -2530 response units) and used as a reference. The analyte protein (Ran, Nup153 ZnF2, or Ub) was injected in duplicate over the four flow cells at increasing concentrations (100 l/min for 30 s) in one of two buffers, either 20 mM HEPES, pH 7.3, 110 mM KOAc, 5 mM NaOAc, 2 mM MgOAc, 0.5 mM EGTA, 1 mM DTT, 1 g/ml aprotinin and leupeptin, 1% glycerol, 0.005% P20, 200 g/ml bovine serum albumin or 25 mM Tris, pH 7.0, 5 mM MgCl 2 , 2 mM DTT, 10 mM ZnCl 2 , 0.01% P20, 200 g/ml bovine serum albumin. When GST-Ran proteins were immobilized, 1 mM GTP or GDP (corresponding to the loaded nucleotide) was also added to the buffer. Background binding to GST surfaces (negligible in all cases) was subtracted before calculating equi-librium dissociation constants. The equilibrium response data were then fit to either a 1:1 or a two independent sites model as indicated under "Results" (60). In some instances, duplicate injections did not overlay exactly (e.g. Fig. 1C) indicating a decrease in response that was likely due to a loss of ZnF activity over time. To account for this, each set of analyte concentration was fit individually to obtain K D information (61).
Importin ␤ Binding Assay-Binding assays were performed at 23°C in binding buffer (25 mM Tris, pH 7.0, 5 mM imidazole, 5 mM MgCl 2 , 1 mM DTT, 10 M ZnCl 2 , 0.005% Tween 20, 200 g/ml bovine serum albumin, 10 g/ml aprotinin and leupeptin). Ni-NTA resin (10 l per reaction) was equilibrated 3 times with 1 ml of binding buffer. For each reaction, 120 pmol of His-tagged importin ␤ fragments 1-462 or 45-462 were immobilized on the beads, washed, and incubated with 5 pmol of GST-RanQ69L. The appropriate nucleotide was maintained in the binding buffer at 1 mM. The matrices were then washed three times with buffer containing nucleotide. Bound proteins were eluted in SDS sample buffer, separated by SDS-PAGE, and visualized in immunoblots with antibodies directed against the His tag on the importin ␤ fragments (Qiagen) or Ran (Transduction Laboratories).

Individual Nup153
ZnF Modules Bind Ran-RanGDP interacts with the zinc finger regions of two nuclear pore proteins, Nup153 and Nup358 (32,34). These Ran binding regions contain 4 -8 ZnF modules connected by linkers that vary in sequence and length (28 -42 amino acids, Fig. 1, A and B). Previous studies have not defined a minimal Ran binding element, although individual zinc fingers have been shown to be sufficient for COPI association and to exert dominant negative inhibition of nuclear envelope breakdown, albeit with weaker activities than the intact tandem array of zinc fingers (19). These experiments suggested that Nup ZnFs can act as functional units, and we therefore began by testing whether individual Nup153 ZnF modules could bind specifically to RanGDP.
GST-Nup153 ZnF fusion proteins (supplemental Fig. S2) were immobilized onto biosensor chips and purified RanGDP was injected over each surface. All four Nup153 ZnF modules bound to RanGDP (Fig. 1, C and D), and the interactions were specific because RanGDP did not bind to a mutant version of the second zinc finger (ZnF2mut, Fig. 1C, inset, and described in detail in Fig. 5) or to GST alone (data not shown).
In all cases, equilibrium binding and complete dissociation were achieved within seconds ( Fig. 1C and data not shown). The equilibrium binding phases of the experiment were therefore analyzed to obtain equilibrium binding constants. The data set shown in Fig. 1D Table 1).
Having established that a single zinc finger motif can bind to RanGDP, we next wished to determine how the native, tandem context of these ZnF modules influences Ran binding. To do so, we immobilized a GST fusion protein (4ZnF) encompassing all four zinc fingers and residues that link them (amino acids 658 -1002, Fig. 1A) on a biosensor chip. ZnF2 and ZnF3 were immobilized for direct comparison. Measurements of RanGDP binding to the full domain fit well to a two-independent site model, with one class of sites similar in affinity to ZnFs 1 and 2 and a second class similar to ZnFs 3 and 4 (Fig. 1E). Relative to the single zinc fingers, Ran bound to the 4ZnF fragment with a higher stoichiometry, consistent with this region containing multiple zinc finger binding sites. We therefore conclude that: 1) all four Nup153 ZnFs can bind RanGDP; 2) individual ZnF motifs can function as discrete Ran binding modules; 3) absolute RanGDP binding affinities vary with ZnF sequence, with the first two Nup153 ZnFs binding ϳ6-fold more tightly than the final two; and 4) when in their native, tandem context, the Nup zinc fingers bind Ran with little cooperativity or contribution from linker regions.
NMR Studies of a Nup Zinc Finger-To gain insight into the molecular basis for RanGDP binding, we determined the three-dimensional structure of the second ZnF module of human Nup153 (Nup153 ZnF2). Using standard 1 H/ 13 C/ 15 N triple resonance heteronuclear NMR experiments, nearly complete resonance and stereo-specific assign-  Table 1. E, biosensor isotherms for Ran binding to ZnF2 (f), ZnF3 (OE), and the full domain with four zinc fingers, 4ZnF (؉). Binding is graphed as a function of the relative amount of Ran bound. The data for the 4ZnF domain fit well to a two-independent site model and the two K D values are indicated as well as the K D values calculated for ZnF2 and ZnF3 in this experiment. a The order of individual zinc fingers of Nup153 within the zinc finger domain are denoted by number (e.g. ZnF1 is the first zinc finger of the domain). Zinc finger mutations have been represented by the zinc finger changed followed by the amino acids the mutations resulted in (e.g. ZnF2-TF indicates that the 2nd zinc finger has been mutated from endogenous residues (leucine, valine) to threonine, phenylalanine (TF). b Superscripts denote starting position of the zinc finger within the relevant protein. Bold lettering represents amino acids mutated from the wild-type sequence. c n represents the number of individual experiments performed to test the interaction between zinc finger and analyte. d Unless otherwise noted, K D values were measured in Biacore biosensor experiments and error determined by the standard deviation between n experiments. e Dissociation constants and error were determined from a statistical fit of a single binding isotherm derived from duplicate measurements at ten different Ran or Ub concentrations (0 -99 M for ZnF2-TF, 0 -709 M for ZnF2-TFM, or 0 -828 M for Npl4). f K D with error could only be obtained for one of two experiments at 5700 Ϯ 100 M. The second experiment was measured at Ͼ2000 M with no measurable error. ments were obtained for Nup153 ZnF2, except for the amide protons of the first two non-native N-terminal residues (GH) and the H␣ of Lys 23 , which resides under the water resonance. As expected, the C␤ shifts for the coordinating cysteine residues indicated that those side chains coordinated the single zinc ion. As discussed below, two residues located on either side of the single tryptophan (Asn 16 and Lys 30 ) also exhibited notable ring current shifts.
Initial structural refinements of the zinc finger used exclusively NOE data and provided the overall domain fold in the absence of zinc coordination restraints. Torsion angle dynamics refinements within CYANA were then completed by adding additional restraints, including stereospecific H␤ proton assignments, 1, , and torsion angles, hydrogen bonds, and idealized zinc geometry (based on the Npl4 zinc finger (24)). The final refinement included 362 NOEs, 25 dihedral restraints, 20 stereo assignments, 14 zinc coordination-geometry restraints, and 12 hydrogen bonding restraints (see Table SI). The first four (1-4) and the final (31) residues were disordered, but the remaining residues were all well defined, and the final ensemble of NMR models was of high quality, with low average target functions, good backbone geometry and stereochemistry, and small root mean square deviations from the mean structure (0.11 Å for backbone atoms, 0.64 Å for all heavy atoms, see Fig. 2A, Table SI, and Protein Data Bank code 2GQE).

JOURNAL OF BIOLOGICAL CHEMISTRY 17095
␤-sheet hydrogen bonding patterns (N16O-W7H, N16H-W7O, and V14O-C 9H ), whereas analogous hydrogen bonds were only formed at one end of the second hairpin (K30H-I21O and C23H-T28O). The two ␤-hairpins surround a small hydrophobic core that consists of the Trp 7 indole ring capped at either end by the Asn 16 and Lys 30 side chains. Asn 16 and Trp 7 are nearly invariant, whereas position 30 (Lys 30 here) is typically a Lys or an Arg. The Trp 7 , Asn 16 , and Lys 30 core residues form an extensive hydrogen bonding network that stabilizes the fold (Fig. 2C). Not unexpectedly, the backbone conformation of Nup153 ZnF2 is very similar to other RanBP/NZF modules with published structures (Fig. 2D) (24,26,27,62).
Mapping the Ran Binding Surface of Nup153 ZnF2-To understand how Nup153 ZnF2 creates Ran binding specificity within this shared architectural framework, the Ran binding surface on Nup153 ZnF2 was mapped using NMR chemical shift perturbation experiments (Fig. 4). Chemical shift changes revealed that the primary Ran binding site on Nup153 ZnF2 was centered about Cys 12 , Leu 13 , Val 14 , and Ala 25 , with smaller but still significant shifts in surrounding residues (Thr 6 , Trp 7 , Cys 9 , Val 24 , and Cys 26 ). All but two of the Nup153 ZnF backbone amides in the complex were in fast exchange, as is typical for binding interactions with micromolar dissociation constants. The Leu 13 and Val 14 amides in the center of the Ran binding site were in intermediate exchange, which presumably reflects their unusually large chemical shift changes upon Ran binding.
The Nup153 ZnF2 amino acids implicated in Ran binding are clustered and exposed on one edge of the zinc coordination site (Fig. 4, B and C). The position of this interaction site is essentially identical to the ubiquitin binding site on Npl4 NZF ( Fig.  4C and Ref. 24). In particular, the three amino acid positions that form the core of the Npl4 NZF ubiquitin binding site (Thr 13 , Phe 14 , and Met 25 ) also form the core of the Nup153 ZnF2 Ran binding site (Leu 13 , Val 14 , and Ala 25 ). Hence, in both cases the NZF fold serves to cluster residues 13, 14, and 25 into one continuous, exposed recognition surface.
To confirm the importance of the Nup153 ZnF Leu 13 ,Val 14 ,Ala 25 motif for Ran binding, we initially tested the effects of replacing this motif with the analogous Thr 13 ,Phe 14 ,Met 25 motif from Npl4 NZF. As expected, the triple substitution mutation abrogated RanGDP binding (Fig. 5A). Indeed, even the L13T,V14F double mutation alone reduced the RanGDP binding affinity more than 300-fold (Table 1, hZnF2-TF). To monitor the integrity of the hybrid Nup153 ZnF2-TFM, we tested this mutant for ubiquitin binding, a property known to be conferred by these amino acids within an NZF backbone (25), and found that binding affinity of Nup153 ZnF2-TFM for ubiquitin was only 2-fold weaker than wild-type Npl4 NZF (K D ϭ 206 versus 122 M; Table 1). In the converse experiment, substituting the Leu 13 ,Val 14 ,Ala 25 motif into the Npl4 NZF scaffold enhanced RanGDP binding activity, although the binding was weak (K D ϭ 350 M; Fig. 5B and Table  1). We therefore conclude that, although surrounding residues can clearly modulate binding affinities, the "LVA" motif constitutes the primary determinant of Ran binding within a highly conserved NZF structural scaffold.
Interaction with Nup153 ZnFs Is Minimally Affected by the Nucleotide Bound by Ran-The zinc finger domains of Nup153 and Nup358 have previously been characterized as binding specifically to the GDP form of Ran (32,34) and this activity has been broadly annotated in genomic data bases. With a quantitative assay in hand, we compared the relative binding affinities of RanGDP and RanGTP for Nup153 ZnF2. GST-Ran was first loaded with either GDP or GTP and immobilized on the biosensor chip. Initial experiments employed the mutant RanQ69L, which is defective in GTP hydrolysis, to ensure the stability of the loaded nucleotide (65). Surprisingly, Nup153 ZnF2 showed only a slight binding preference for RanQ69L-GDP (K D ϭ 7 Ϯ 1 M) over RanQ69L-GTP (K D ϭ 12 Ϯ 1 M) ( Fig. 6A and Table 2). Nup153 ZnF2 also bound with similar affinities to the GDP and GTP forms of wild-type Ran (K D ϭ 8 Ϯ 0.4 M and K D ϭ 9 Ϯ 1, respectively; Table 2). Additional experiments with ZnF2, ZnF3, and the full 4ZnF fragment immobilized and tested for binding to Ran loaded with GDP or GTP were consistent with this trend. Neither ZnF3 nor 4ZnF showed significant preference for RanGDP over RanGTP (data not shown). Also of note, similar equilibrium dissociation constants were obtained for all Nup153 ZnF2/RanGDP interactions regardless of which protein was immobilized or whether the wild-type or Q69L RanGDP proteins were used. These similarities helped to confirm the reliability of the measurements and established that the Q69L mutation did not influence Nup153 ZnF2 binding.
To confirm that our RanGTP and RanGDP preparations were loaded with the appropriate nucleotide, we took advantage of the well characterized interaction between RanGTP and importin ␤. Importin ␤ binds RanGTP with very high affinity (K D ϭ 140 pM) (66), whereas its affinity for RanGDP is much lower (K D ϳ 5-10 M) (67). This interaction can be recapitulated with a fragment containing the first 462 amino acids of importin ␤, and removal of the first 44 amino acids from this fragment eliminates Ran binding (40,68). Interaction assays were performed by immobilizing either His-tagged importin ␤-(1-462) or importin ␤-(45-462) (negative control) on Ni-NTA resin. Immobilized importin ␤ fragments were then incubated with purified GST-RanQ69L, previously loaded with either GDP or GTP. The appropriate nucleotide was maintained at 1 mM in each binding assay, as in the corresponding Nup153 ZnF2 biosensor assays. The resin was washed and bound proteins were eluted and analyzed by immunoblot. As expected, binding was observed with only the longer importin ␤-(1-462) construct and only for GTP-loaded RanQ69L (Fig.  6B). These experiments confirmed that our experimental procedures produced Ran proteins that were loaded predominantly with the expected nucleotides. Our binding experiments therefore imply that Nup153 ZnF recognition is affected only modestly, if at all, by the Ran nucleotide state.

DISCUSSION
The structure of Nup153 ZnF2 demonstrated that Nup zinc fingers adopt the minimal RanBP2/NZF zinc ribbon fold first seen for the Npl4 NZF protein (24). As illustrated in Fig. 2D, it is now clear that RanBP2/NZF modules adopt very similar backbone conformations. However, the functional utility of the RanBP2/NZF classification has been limited by a lack of information on the specific binding interactions of the hundreds of different proteins that contain this structurally conserved module.
The present study demonstrates that a subset of RanBP2/ NZF modules can bind Ran, and defines a sequence motif that can be used to predict whether or not individual RanBP2/NZF fingers will bind Ran. Specifically, we have shown that three key residues: Leu 14 , Val 15 , and Ala 25 , form the primary Ran binding site on Nup153 ZnF2. This LVA motif is not present in most RanBP2/NZF sequences, but is highly overrepresented in the Nup ZnF subfamily (supplemental Fig. S1). The three LVA residues also exhibit significant co-variation (68% of NZF modules with LV have an Ala at position 25), suggesting that they are linked functionally. Moreover, double or triple mutations within the LVA motif abrogate Ran binding, and Ran binding activity is created when the LVA motif is transferred into a heterologous NZF module. The hybrid zinc finger (Npl4-LVA) does not bind Ran as tightly as the wild-type Nup zinc fingers, however, indicating that elements outside the LVA motif influence binding. This is also evidenced by the fact that Nup153 ZnF2 and ZnF4 bind Ran with different affinities (6 versus 32 M), even though both contain the LVA motif. Finally, we note that conservative changes within the LVA motif can clearly be tolerated as Nup153 ZnF1 has a V14L substitution and ZnF3 has L13C and A25S substitutions.
Our data indicate that most if not all RanBP2/NZF modules that contain the LVA motif will bind Ran with micromolar affinities. This predicts RanGDP binding for at least the six of eight RanBP2/NZF repeats in human Nup358 that contain LVA motifs. This prediction is supported by preliminary biosensor experiments showing that the third ZnF module from Nup358 also bound RanGDP (K D ϭ 51 M).
Comparison of Ran binding to individual zinc fingers versus a fragment containing the native, tandem array of zinc fingers led us to conclude that linker regions do not contribute significantly to the interface with Ran, either by facilitating cooperative interactions or by providing additional contact sites, as was recently observed for the N-terminal Vps36 NZF in which downstream residues were found important for its interaction with Vps28 (62). Although the Nup ZnFs appear to operate as independent Ran binding sites, the presence of these arrayed sites within both Nup153 and Nup358 could, in principle, cre-   41)). Ran and importin ␤ were detected by immunoblot using anti-Ran (upper panels) and penta-His antibodies (lower panels), respectively. a Dissociation constant is the mean of two independent measurements and the error is the range in the two measurements. b Dissociation constant and error were estimated from a statistical fit of a single binding isotherm derived from duplicate measurements at 12 different hZnF2.

Nup153 ZnF-Ran Interactions
ate high avidity, multivalent Ran binding platforms on either side of the nuclear pore if multiple copies of Ran proteins are present in stable complexes. Whether or not they contribute to the avidity of interactions at the pore, the 32-48 predicted Ran-binding ZnFs on each face of the pore (due to its 8-fold symmetry) may ensure that Ran, in either GTP or GDP bound form, is concentrated in the vicinity of the nuclear pore complex, poised to work in conjunction with other ZnF-binding proteins, to participate in regulation of transport receptor-cargo interactions, and/or to interact with regulatory factors, such as RanGEF/RCC1, or the GTPase activating protein, RanGAP. A global search for NZF motifs containing an LVA signature revealed that this sub-group is restricted to proteins known or annotated as nucleoporins. Notably, two subspecies of the intracellular parasite, Theileria, express proteins (CAI75897, XP_762801) with RanBP2/NZF modules that have the sequence "LVC" in these key positions. It would be of interest to know if this related motif also supports Ran binding because these proteins additionally contain homology to the guanine exchange factor RanGEF/RCC1, suggesting that the zinc finger module may play a role in delivering Ran for nucleotide exchange, in this context using domains within the same protein.
Interestingly, Palmenberg and colleagues (69) recently reported a zinc finger-containing protein (L) in encephalomyelitis virus, a member of the picornavirus family, which is able to bind Ran. Mutation of key structural residues in the zinc finger reduced binding to Ran. The NMR structure of the zinc finger from the L protein of a related virus, Mengovirus, was recently solved (PDB code 2BAI) and is structurally distinct from the Nup ZnF. Overlay of the two modules does not clearly align a binding platform that is analogous to LVA, indicating that this ZnF/Ran interaction may not closely mimic the endogenous ZnF/Ran interaction reported here.
Although this viral ZnF/Ran interaction was found to be independent of whether Ran was bound to GTP or GDP, in light of a previous study on Nup153 (32), we were surprised to find that Nup153 ZnFs recognize Ran in both its GTP and GDP conformations, with only a slight preference for the GDPbound form. The interaction between Ran and the Nup153 ZnF domain was previously assessed relative to the interaction between Ran and importin ␤, and the high affinity of this latter interaction may have "masked" a significant, but appreciably weaker, Nup153/RanGTP interaction (32). Nonetheless, because our assay for nucleotide loading is qualitative, it is difficult to reach an indisputable conclusion here. A definitive understanding of the influence of nucleotide awaits future structural studies of the Ran⅐Nup-ZnF complex, which will reveal the important determinants within Ran for this partnership and whether they include regions modulated by nucleotide association.
More information about the Ran side of the Nup ZnF:Ran interface may also reveal whether Nup zinc fingers can bind to other members of the extensive family of small GTPases. Our preliminary data indicate that Nup153 ZnF does not bind Arf1, one such related GTPase (data not shown), but this question still needs to be addressed more thoroughly. Given the range of NZF modules and the extent of the Ras superfamily, it would not be surprising if other NZF family members bind other Rasrelated proteins.