Structure of a Yeast Dyn2-Nup159 Complex and Molecular Basis for Dynein Light Chain-Nuclear Pore Interaction*

Background: Dyn2 and Nup159 interact to induce oligomerization of the Nup82 cytoplasmic fibril complex. Results: Dyn2 homodimers symmetrically bind Nup159 target sites through an antiparallel β-strand interaction. Conclusion: Dyn2 homodimers bind diverse, arrayed Nup159 target sites, locally directing Nup159 homodimerization. Significance: Learning how nucleoporins interact to form fibril architecture is crucial for understanding how the nuclear pore complex gates traffic. The nuclear pore complex gates nucleocytoplasmic transport through a massive, eight-fold symmetric channel capped by a nucleoplasmic basket and structurally unique, cytoplasmic fibrils whose tentacles bind and regulate asymmetric traffic. The conserved Nup82 complex, composed of Nsp1, Nup82, and Nup159, forms the unique cytoplasmic fibrils that regulate mRNA nuclear export. Although the nuclear pore complex plays a fundamental, conserved role in nuclear trafficking, structural information about the cytoplasmic fibrils is limited. Here, we investigate the structural and biochemical interactions between Saccharomyces cerevisiae Nup159 and the nucleoporin, Dyn2. We find that Dyn2 is predominantly a homodimer and binds arrayed sites on Nup159, promoting the Nup159 parallel homodimerization. We present the first structure of Dyn2, determined at 1.85 Å resolution, complexed with a Nup159 target peptide. Dyn2 resembles homologous metazoan dynein light chains, forming homodimeric composite substrate binding sites that engage two independent 10-residue target motifs, imparting a β-strand structure to each peptide via antiparallel extension of the Dyn2 core β-sandwich. Dyn2 recognizes a highly conserved QT motif while allowing sequence plasticity in the flanking residues of the peptide. Isothermal titration calorimetric analysis of the comparative binding of Dyn2 to two Nup159 target sites shows similar affinities (18 and 13 μm), but divergent thermal binding modes. Dyn2 homodimers are arrayed in the crystal lattice, likely mimicking the arrayed architecture of Dyn2 on the Nup159 multivalent binding sites. Crystallographic interdimer interactions potentially reflect a cooperative basis for Dyn2-Nup159 complex formation. Our data highlight the determinants that mediate oligomerization of the Nup82 complex and promote a directed, elongated cytoplasmic fibril architecture.

The Saccharomyces cerevisiae nuclear pore complex (NPC) 2 is a 66-MDa structure composed of ϳ30 different proteins that embed in the nuclear envelope and facilitate transport across this barrier (1). NPC proteins are highly conserved in function and sequence across eukaryotes and carry out biologically conserved functions: mRNA export into the cytoplasm and gated transport of specific proteins into and out of the nucleus.
The proteins that make up this highly coordinated and complex structure form an eight-fold symmetrical pore from a limited number of structural folds (2). The types of domains in the nuclear pore proteins, or nucleoporins (Nups), are primarily ␣-solenoids, ␤-propellers, Phe-Gly rich repeats, coiled-coil domains, and transmembrane domains (3,4). Transmembrane domains traverse the double nuclear envelope membrane and underlie the NPC core topology and biogenesis (5,6). Phe-Gly repeats are primarily concentrated in the core interior where they function as a physical or entropic barrier to entering proteins while reversibly binding nuclear transport receptors and selectively allowing their passage (1,7). On either side of the core, asymmetrically distributed elements are positioned to facilitate asymmetric, unidirectional transport. The nucleoplasmic side of the NPC contains proteins tethered into a basket-like structure that protrudes 95 nm into the nucleus, potentially serving as a molecular checkpoint for pre-mRNA before it exits the nucleus (8,9). On the cytoplasmic surface, NPC fibrils stretch 50 nm into the cytoplasm (8). Cytoplasmic fibrils are primarily composed of nucleoporins from the Nup82 complex that bind translation initiation factors and mRNA export machinery (10). The Nup82 complex consists of Nup82, Nup159, and Nsp1 that work with Nup116, Nup42, Gle1, and Nup100 to mediate mRNA export in concert with the mRNA nuclear export receptor Mex67 and the DEAD box RNA helicase Dbp5 (4,(11)(12)(13).
Nup159 is a prime component of the Nup82 complex and plays a directed role in coordinating nucleoporins involved in mRNA export rather than protein trafficking between the cytoplasm and nucleus (14). Nup159 has an extended multicomponent architecture that facilitates its roles in mRNA export, as well as filament localization in the NPC structure (14 -17). The Nup159 N-terminal domain constitutes a seven-bladed ␤-propeller that extends into the cytoplasm and mediates Dbp5 binding (18). Deletion of the Nup159 N-terminal domain results in a temperature-sensitive phenotype, lethal at 37°C and hallmarked by Dbp5 mislocalization and constitutive mRNA export defects at 23°C (14,16). The Nup159 central 700 amino acids form a Phe-Gly-rich repeat domain. C-terminal to the Phe-Gly-rich repeats is a 100-amino acid region termed the dynein light chain-interacting domain (DID) that uses a pentameric array of dynein light chain binding motifs to bind the yeast dynein light chain Dyn2 (11). C-terminal to the DID, Nup159 contains a predicted helical region (19) that is essential for the stability and localization of Nup159 on the NPC and has recently been shown to form a heterotrimeric structure with Nup82 and Nup116 (14,16,20). Higher order oligomerization of the Nup82 complex requires both the Nup159 DID region as well as Dyn2 (11). The functional role of a dynein light chain at the nuclear pore is independent of its role in the cytoplasmic dynein microtubule motor complex (11).
The dynein light chain is a promiscuous protein, involved in a diversity of protein target interactions, only a subset of which involve binding to the cytoplasmic dynein microtubule motor complex. In S. cerevisiae, Dyn2 interacts with the dynein intermediate chain, Pac11, through tandem canonical 10 -12-residue stretches, each containing a conserved QT motif (21). Similar Dyn2 binding motifs are arrayed in Nup159, where class averaged electron microscopy of the Nup159 DID region saturated with Dyn2 showed a stack of five densities like beads on a string, effectively mediating the Nup159 dimerization (11). As with Dyn2, the higher dynein light chain orthologs, dynein light chain 8 (LC8) and dynein light chain (DYNLL), bind partners outside of the dynein motor complex including the signaling molecules neuronal NOS and Pak1, the apoptosis regulator Bim/Bmf, and the mRNA localization protein Swallow (22)(23)(24). Given the diverse set of dynein light chain partners, it has been postulated that the dynein light chain functionally serves as a dimerization machine. This role correlates with the structures of higher Dyn2 orthologs that show dynein light chains complexed 2:2 with a variety of target peptides (23,(25)(26)(27)(28).
Although studies to date have biophysically characterized dynein light chains from Drosophila, rat, and human, molecular details of the S. cerevisiae dynein light chain have remained outstanding. S. cerevisiae is a leading model system for biophysical, biochemical, and genetic investigations of the NPC and the cytoplasmic dynein motor complex. To further our molecular understanding of Dyn2 and its functional role in the NPC, we determined the x-ray crystal structure of Dyn2 in complex with a Nup159 target site. We couple structural data with gel filtration, multiangle light scattering, and isothermal titration calorimetry to derive a model for the Dyn2-Nup159 interaction and the role of Dyn2 as a dimerization machine.

EXPERIMENTAL PROCEDURES
Cloning and Expression of Full-length Dyn2 from S. cerevisiae-Full-length Dyn2 was cloned from S. cerevisiae S288c into the pGEX-6P2 expression vector (GE Healthcare) using the polymerase chain reaction and BamHI-and EcoRIengineered flanking restriction sites. The Dyn2 insert was sequence-verified against GenBank TM accession NC_001136. pGEX-6P2-Dyn2 was transformed into Escherichia coli BL21 DE3 (pLysS) and grown under ampicillin selection in 6 liters of LB medium at 37°C. At an optical density of 0.8 (600 nm), GST-Dyn2 expression was induced using 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 16 h at 18°C. Cells were harvested by centrifugation at 2100 ϫ g for 10 min at 4°C, and the pellets were resuspended in buffer A (150 ml of 25 mM HEPES, pH 6.8, 300 mM NaCl, and 0.1% ␤-mercaptoethanol) and stored at Ϫ20°C.
Protein Purification-Resuspended cell pellets were thawed and lysed by sonication at 4°C. 0.1 mM phenylmethylsulfonyl fluoride was added to the lysate, and cell debris was pelleted by centrifugation at 23,000 ϫ g for 45 min. Supernatant was loaded onto a 5-ml glutathione-Sepharose Fast Flow affinity column (GE Healthcare). GST-tagged Dyn2 was eluted from the glutathione column with 100 ml of 3 mM glutathione, pH 8.0, in buffer A. The GST-Dyn2 eluate was exchanged into buffer B (25 mM HEPES, pH 6.8, and 0.1% ␤-mercaptoethanol) using an Amicon Ultra 10-kDa spin concentrator (Millipore) and incubated for 16 h with PreScission protease (GE Healthcare). The cleaved protein was loaded onto an SP Sepharose Fast Flow column (GE Healthcare) and eluted over a linear 0 -1 M NaCl gradient in buffer B. Dyn2 peak fractions were pooled and exchanged into 50 mM NaCl, 25 mM HEPES, pH 6.8, and 0.1% ␤-mercaptoethanol using an Amicon Ultra 3-kDa spin concentrator (Millipore) and concentrated to 5 mg/ml, snap-frozen in liquid nitrogen, and stored at Ϫ80°C. All purification procedures were executed at 4°C. The final, purified Dyn2 protein contained an N-terminal GPLGS cloning artifact.
Crystallization-1.0 mM Dyn2 was incubated with 1.5 mM Nup159 pep2 in 50 mM NaCl, 25 mM HEPES, pH 6.8, and 0.1% ␤-mercaptoethanol for 30 min on ice. Crystallization followed the hanging drop protocol using 1 l of the Dyn2-Nup159 pep2 mixture and 1 l of the 1-ml well solution: 0.3 M ammonium acetate, pH 5.5, 5% methyl pentanediol, and 35% polyethylene glycol 4000. Crystals grew at 20°C to 200 ϫ 200 ϫ 600 m over the course of a week. Crystals were transferred to Fomblin oil (Sigma) for cryoprotection and flash-frozen in liquid nitrogen.
Data Collection, Structure Determination, and Refinement-Dyn2-Nup159 pep2 crystals were maintained at 100 K under a cryo-cooled nitrogen stream, and diffraction data were collected using a Rigaku MicroMax 007HF x-ray generator (copper anode, 1.54 Å wavelength), Osmic mirrors, and a Rigaku Saturn 944ϩ charge-coupled device in the University of North Carolina Macromolecular X-Ray Crystallography Core Facility. 0.5°oscillations were collected over 160°from a single crystal. Data were indexed, integrated, and scaled using HKL2000 (29) (see Table 1). The structure was determined using the AutoMR molecular replacement program (PHENIX crystallographic suite (30)) and a modified 2PG1 (26) coordinate file in which a monomeric, apo Drosophila LC8 search model was used. The model was built using AutoBuild (PHENIX) (30) and refined iteratively through manual builds in Coot (31) followed by refinement runs using phenix.refine (PHENIX) (30). Refinement statistics were monitored using a Free R, calculated using 10% of the data, randomly excluded from refinement (32). The final model includes two Dyn2 molecules (chains A and C: residues 7-92), two Nup159 pep2 molecules (chain B: residues 1117-1126; chain D: residues 1116 -1126 with Asn-1116 modeled as alanine), and 217 water molecules.
Size Exclusion Chromatography and Multiangle Light Scattering-100 l of 200 M Dyn2 was injected onto a Wyatt WTC030S5 silicone size exclusion column (for elution of 5,000 -1,250,000-Da proteins) in 50 mM NaCl, 25 mM HEPES, pH 6.8, 0.1% ␤-mercaptoethanol, and 0.2 g/liter sodium azide and passed in tandem through a Wyatt DAWN HELEOS II light scattering instrument and a Wyatt Optilab rEX refractometer. The light scattering and refractive index data were used to calculate the weight-averaged molar mass and the mass fraction in each peak using the Wyatt Astra V software program (Wyatt Technology Corp.) (33).
Isothermal Titration Microcalorimetry-Isothermal titration calorimetry experiments were carried out at 15°C in buffer C: 50 mM NaCl, 25 mM HEPES, pH 6.8, and 0.1% ␤-mercaptoethanol on a MicroCal AutoITC200 (GE Healthcare). Peptides were exchanged into buffer C using G-25 Sephadex quick spin columns (Roche Applied Science). 17 ϫ 2-l injections of 1 mM pep2 or pep4 were automatically injected into 200 l of 100 M Dyn2. The resulting binding isotherms (see Fig. 6, A and B) were analyzed using the Origin 7.0 software package (Origin-Lab) and were fit to a one-site binding model. Experiments were conducted in triplicate and averaged to determine respective mean K D values with standard deviations as shown.

S. cerevisiae Dyn2 Is a member of the Conserved Dynein Light
Chain Family-The dynein light chain, a component of the cytoplasmic dynein motor complex, is highly conserved from yeast to human (Fig. 1A). The dynein light chain is 90% identical across higher eukaryotes ranging from Caenorhabditis elegans to human, with significant identity extending to lower eukaryotes, as exemplified by the 50% identity between S. cerevisiae Dyn2 and Drosophila melanogaster LC8. Across organisms, evidence points to the role of the dynein light chain as a factor that promotes substrate dimerization. Although the dynein light chain is a component of the dynein microtubule motor complex, it is not exclusive to this complex. Recent work has shown that ϳ25% of the S. cerevisiae dynein light chain member, Dyn2, is associated with the nuclear pore complex. Dyn2 associates with the Nup82 cytoplasmic fibril complex, binding to pentavalent motifs arrayed in the Nup159 DID ( Fig.  1, B-D) (11). The Dyn2 binding motifs share a canonical QT motif with variable flanking components. Similar tandem binding sites have recently been mapped in the dynein intermediate chain, Pac11, and shown to mediate Dyn2 interaction (Fig. 1, C and D) (21). To understand the molecular basis of the Dyn2-Nup159 interaction, we cloned Dyn2 from S. cerevisiae (S288c) genomic DNA into the E. coli expression vector pGEX-6P2 and expressed and purified Dyn2 to homogeneity, removing the N-terminal GST tag. Nup159 peptides corresponding to the second and fourth Dyn2 DID binding sites (pep2 and pep4) were synthesized, purified by HPLC chromatography, and verified by mass spectrometry analysis. Pep4 incorporated an N-terminal tyrosine to facilitate concentration determination, whereas pep2 concentration was determined using its endogenous tyrosine.
The Dyn2 Homodimer Forms Two Composite Substrate Binding Sites Using a Central ␤-Sandwich and Flanking ␣2-Helices-To elucidate the structural determinants underlying the Dyn2-Nup159 interaction, we screened mixtures of Dyn2 and Nup159 pep2 and pep4 for co-crystallization. We obtained crystals of the Dyn2-Nup159 pep2 (residues 1116 -1126) complex using a 1:1.5 molar ratio of Dyn2 and Nup159 pep2, respectively. The crystals diffracted to 1.85 Å resolution and belonged to the space group P2 1 2 1 2 1 . We solved the structure by the molecular replacement method using a peptide-free monomeric chain derived from the Drosophila dynein light chain (2PG1) that showed 50% sequence identity with Dyn2 (26). Two Dyn2 chains occupy the asymmetric unit, homodimerized around a noncrystallographic two-fold axis. Clear electron density was evident for two Nup159 pep2 chains, each bound to the Dyn2 homodimer. The R and R free values for the Dyn2-Nup159 pep2 structure are 15.1 and 18.0%, respectively. Crystallographic data and refinement statistics are presented in Table 1.
Dyn2 homodimerizes across a composite central ␤-sandwich ( Fig. 2A). Each ␤-sheet is composed of five ␤-strands arranged in an antiparallel organization: ␤1-␤4-␤5-␤2-␤3Ј in which the final ␤3Ј strand is provided by the homodimeric mate. The prime interface between Dyn2 molecules is mediated by the antiparallel ␤2-␤3Ј strand interaction. Here, the ␤2-␤3Ј and ␤2Ј-␤3 strand interactions encompass the noncrystallographic two-fold operator that relates each Dyn2 molecule. Flanking the central ␤-sandwich, each Dyn2 molecule contributes an ␣1 and ␣2 helix that bridge ␤1 and ␤2. The ␣1-␣2 helix-turn-helix motifs symmetrically pack against the two ␤-sheets that form the central ␤-sandwich. The Dyn2 homodimer symmetrically binds two Nup159 peptides; the basis for the interaction is an extension of each ␤-sheet through an antiparallel strand that is stabilized through buttressing interactions with the neighboring ␣2-helix.
Dyn2 architecture is homologous to other dynein light chain structures determined to date, with the highest structural homology to the human LC8 complexed with a peptide from the protein inhibitor of neuronal nitric oxide synthase (PIN) (Protein Data Bank (PDB) 1CMI), 0.6 Å C␣ r.m.s.d. over 87 residues (47% identity, Fig. 2B), and ranged among dynein light chain structures to 2.3 Å C␣ r.m.s.d. over 81 residues when compared with the dynein light chain structure 1YO3 from Plasmodium falciparum (37% identity) (34,35). The main elements that show structural diversity between Dyn2 and the FIGURE 1. S. cerevisiae Dyn2 is a conserved dynein light chain involved in diverse macromolecular complexes including the nuclear pore complex and cytoplasmic dynein motor complex. A, sequence alignment of 12 dynein light chain family members ranging from S. cerevisiae to human. Residues aligned with 100 and 80% identity are colored green and yellow, respectively. Amino acid numbers and secondary structure elements, based on the S. cerevisiae Dyn2 structure, are shown above the alignment. Residues involved in Dyn2 dimerization and Dyn2-Nup159 pep2 binding are indicated below the alignment by asterisks based on European Molecular Biology Laboratory-European Bioinformatics Institute (EMBL-EBI) PDBe Protein Interfaces, Surface and Assemblies (PISA). Black, Dyn2-Dyn2 chain A-C interactions; red, Dyn2-Nup159_2 chain A-D or chain C-B interactions; blue, Dyn2-Nup159_2 chain A-B or chain C-D interactions. Solvent-accessible (SA) surface area for respective Dyn2 chain C residues is indicated below the alignment, calculated in the presence (black) and absence of the Nup159_2 chain D (gray) using the Accessible Surface Area Analysis tool in CCP4 (41). A. gossypii, Ashbya gossypii; C. dubliniensis, Candida dubliniensis; S. pombe, Schizosaccharomyces pombe; D. discoideum, Dictyostelium discoideum; D. rerio, Danio rerio; X. laevis, Xenopus laevis; R. norvegicus, Rattus norvegicus; P. troglodytes, Pan troglodytes; H. sapiens, Homo sapiens. B, diagram of the nuclear pore complex illustrating the cytoplasmic localization of Nup159 and the Nup82 complex to cytoplasmic fibrils. C, domain architecture of known Dyn2-binding proteins: Nup159 and Pac11. Nup159 is composed of an N-terminal (N-term) ␤-propeller domain involved in Dbp5 binding, central Phe-Gly-rich repeats common to nucleoporins, a DID composed of five QT consensus motifs (residues 1103-1177) with Dyn2 binding activity, and a C-terminal (C-term) region involved in Nup159 anchoring to the nuclear pore complex (11,14,15). Pac11, the yeast dynein intermediate chain, shares architectural similarities with Nup159, composed of an N-terminal coiled-coil domain, tandem Dyn2 QT binding motifs, and a C-terminal WD-40 repeat domain, predicted to be a ␤-propeller. D, sequence alignment of the Dyn2 binding motifs from Nup159 and Pac11 highlighting the invariant QT motif. Nup159_2 secondary structure is indicated above the alignment. Nup159_2 residues involved in Dyn2 binding are indicated by asterisks below the alignment, as is solvent-accessible surface area, calculated in the presence (black) and absence of Dyn2 chains (gray).
Drosophila LC8 structure (1CMI) are restricted to loop regions, specifically the ␣1-␣2 loop, the ␤3-␤4 loop, and the ␤4-␤5 loop. The core secondary structure elements of the domain show little plasticity. Diversity of allowable residues in the target peptide N-terminal to the canonical QT motif in turn shows structural diversity in the target ␤-strand backbone bound to the dynein light chain, as shown in the overlay of the Dyn2-Nup159 peptide structure with the human LC8-PIN peptide structure (Fig. 2B, supplemental Fig. 1).
The Dyn2 homodimerization interface involves an extensive hydrophobic and electrostatic interface that buries ϳ940 Å 2 of solvent-accessible surface area on each Dyn2 molecule. Core ␤-strand-␤-strand hydrogen-bonding networks extend the antiparallel sheets across homodimeric mates (␤2-␤3Јand ␤2Ј-␤3), augmented through additional backbone-side chain electrostatic interactions as well as van der Waals contacts between side chains (Fig. 3A). Helix ␣2Ј packs against the ␤3 strand and buttresses the dimerization interface through the use of charged side chains, primarily Glu-38Ј and Lys-46Ј, that afford van der Waals contacts as well as hydrogen bonding to ␤3 residues Asn-64 and Thr-70, respectively (Fig. 3, A and B). Overall, the homodimerization interface involves a pseudo-symmetric set of reciprocal interactions involving conserved residues (Figs. 1A and 3A).
Dyn2 Exists as a Multimer in Solution-Although the crystal structure of the Dyn2-Nup159 pep2 complex showed Dyn2 in a homodimeric state, we wanted to determine whether this dimeric form existed in solution in the absence of bound peptide. To determine the oligomeric state of Dyn2, we utilized size exclusion chromatography coupled with multiangle light scattering. We analyzed the elution and mass profiles of purified Dyn2 injected at an initial concentration of 200 M. The Dyn2 elution profile contained three main peaks with masses respectively calculated at 25.8, 50.5, and 87.8 kDa. On average, 87% of the eluted mass fraction was in the 25.8-kDa peak (Fig. 3C). The theoretical calculated molecular mass of our Dyn2 construct is 10,852 Da. Thus, under the conditions analyzed, peptide-free Dyn2 was found primarily as a homodimer with the remaining population in higher order oligomeric states.
The Dyn2 Homodimer Binds Parallel Nup159 Peptides Using a Conserved Composite Binding Site-Nup159 contains a pentameric array of Dyn2 binding sites (11). In the structure we present here, the Dyn2 homodimer is complexed with two Nup159 peptides corresponding to the second Dyn2 binding site in the Nup159 DID. The Nup159 peptide binds in a conserved pocket formed at the Dyn2 homodimer interface, consisting of both hydrophobic and charged residues (Fig. 4B). Nup159 pep2 (chain B) buries 911 Å 2 of solvent-accessible surface area, whereas each of the Dyn2 molecules bury 512 and 129 Å 2 , for a collective 641 Å 2 of solvent-accessible surface area buried at a single Nup159 pep2 binding site. 16 of the 25 Dyn2 residues involved in Nup159 peptide binding are 100% invariant across the 12 species shown in Fig. 1A, and 22 of the 25 are at least 80% invariant across these species (Fig. 4A). In Nup159, the glutamine (Gln-1123) and threonine (Thr-1124) that constitute signature dynein light chain binding determinants bind to the periphery of the peptide binding cleft in an area of high dynein light chain conservation. Analysis of the Dyn2 electrostatic surface shows that the peptide binding cleft is composed of mixed charges near the peptide N-terminal region, whereas positive charges dominate the electrostatic potential at the C-terminal region of the peptide. Key salt bridges in the complex include interactions between the invariant Dyn2 Lys-12 and Nup159 Asp-1125 as well as Dyn2 Glu-38Ј and Nup159 Gln-1123. The Nup159 peptides form a ␤-strand interaction, extending the central ␤-sheets formed by Dyn2 homodimerization. The Nup159 ␤-strand runs antiparallel to the Dyn2 ␤3 strand and extends across 7 residues, terminating at the glutamine, Gln-1123, that composes the QT motif (Figs. 4 and 5). The Nup159 Gln-1123 side chain forms a network of hydrogen bonds with the start of the neighboring Dyn2 ␣2Ј helix, capping the end through interactions with the Arg-39Ј backbone amine as well as one of the Glu-38Ј side chain carboxyl oxygens. In addition, the Gln-1123 side chain forms a hydrogen bond with the Phe-65 backbone amine on ␤3. The Gln-1123 backbone is stabilized through a hydrogen bond to the Dyn2 Tyr-78 hydroxyl group (Fig. 5, A-C). Thr-1124 from the Nup159 QT motif forms extensive contacts with Dyn2 Phe-65, engaging the Phe-65 backbone carbonyl and amine through hydrogen bonds from its own backbone amine and side chain hydroxyl group. The Thr-1124 side chain ␥C also forms van der Waals contacts with the Phe-65 benzene ring. Preceding the QT motif, the 6 N-terminal Nup159 residues primarily use an antiparallel ␤-strand-␤-strand hydrogen bond network as well as van der Waals contacts to bind the conserved Dyn2 groove, indicative of the highly variable composition accepted in dynein light is the integrated intensity of the ith reflection with the Miller Index h and ͗I(h)͘ is the average over Friedel and symmetry equivalents. b R value ϭ ͚(͉F obs ͉ Ϫ k͉F calc ͉)/͚͉F obs ͉. c R free is calculated using a 10% subset of the data that are removed randomly from the original data and excluded from refinement. chain targets. Overall, the Nup159 pep2-Dyn2 interface is mediated by extensive hydrogen bonding and van der Waals contacts involving 10 residues in the Nup159 peptide. All Nup159 residues modeled contact one Dyn2 protomer, and 6 of these 10 residues make additional contacts with the Dyn2Ј homodimeric mate, indicating that high affinity Dyn2-substrate recognition is mediated via Dyn2 dimerization. The Nup159 pep2-Dyn2 interface buries 1565 Å 2 of solvent-accessible surface area at each binding site. The two Nup159 peptides bound to the Dyn2 homodimer run parallel to each other, related by a two-fold symmetry axis, with their midpoints separated by ϳ20 Å. Nup159 DID Sites Two and Four, Respectively, Bind Dyn2 with 17.9 and 13.1 M Affinity, Using Differential Thermal Binding Modes-To determine the affinities between Dyn2 and the five Nup159 Dyn2 binding sites in the DID, we synthesized the respective peptides and performed isothermal titration calorimetry, titrating peptides into the calorimeter cell containing Dyn2. Each Nup159 peptide binding experiment was performed in triplicate, and the fitted values were averaged. Each individual binding experiment was best fit to a one-site model (using the Dyn2 monomer concentration) (33). Pep2 showed an endothermic isotherm (Fig. 6A), whereas pep4 showed an exothermic isotherm (Fig. 6B). The experimentally determined affinities between Dyn2 and Nup159 pep2 and pep4 are shown in Fig. 6 and have K D values equal to 17.9 and 13.1 M, respectively. Pep3 did not show sufficient signal to noise and was not soluble at the concentrations needed to determine binding accurately. Pep1 and Pep5 are highly hydrophobic, and once solubilized, failed to show binding to Dyn2 as determined using isothermal titration calorimetry (data not shown). This may be due to the weaker binding affinities for these peptides as was qualitatively shown in the aforementioned PepScan assay (11) or due to a folded/aggregated state that precluded Dyn2 from binding. Structure of Dyn2-Nup159 pep2 complex shows a quaternary complex composed of a Dyn2 dimer, bound to two Nup159 peptides through parallel, composite ␤-sheets. A, diagram of the Dyn2-Nup159 complex. Dyn2 chain A is shown in orange (␣-helices) and dark blue (␤-strands), and Dyn2 chain C is shown in beige (␣-helices) and light blue (␤-strands). The two-fold noncrystallographic symmetry operator that relates the Dyn2 and Nup159 chains in the asymmetric unit is indicated about the z axis. The image at right shows the complex after a 90°rotation about the y axis. B, the complex as shown in the two orientations in A, with Dyn2 chain A superimposed on the human dynein light chain, LC8 (light green) bound to a PIN peptide (yellow) (PDB 1CMI) after a least squares fit with an r.m.s.d. of 0.6 Å over 87 aligned C␣ atoms (34). Helices are shown in cylindrical format. Structural differences between Dyn2 and human LC8 are indicated by red arrows and are dominated by loop regions as well as the bound peptides. MAY 4, 2012 • VOLUME 287 • NUMBER 19

Structure of Dynein LC-Nup159 Complex
Translational Arrangement of Dyn2 Homodimer Facilitates Contiguous Binding to Arrayed QT Motifs-The arrangement of delineated QT motifs in Nup159 is nearly contiguous, sepa-rated by one or two amino acids except for a tentative QT region linking sites two and three that showed no Dyn2 binding activity in the previously mentioned PepScan assay (11). In the same investigation, electron microscopy of the Dyn2-Nup159 DID complex showed five densities arranged like beads on a string, leading the authors to propose a model in which five Dyn2 dimers bound parallel Nup159 DID arrays. Stelter et al. (11) modeled the bound Dyn2 dimers in a translational array. In the P2 1 2 1 2 1 lattice presented here, we note a translational arrangement of Dyn2 dimers in the crystal lattice, which supports the Stelter et al. (11) Dyn2-Nup159 complex model. As shown in Fig. 7, Dyn2-Nup159 pep2 complexes are translationally arranged in the crystal, with a 34 Å translational component approximately collinear to the Nup159 peptide, effectively placing the C terminus of one Nup159 peptide proximal to the N terminus of the neighboring Nup159 peptide. Five Dyn2 dimers in this crystal lattice span 170 Å, on par with the 20-nm filaments observed by Stelter et al. (11) in electron micrographs of the Dyn2-Nup159 DID complex.

DISCUSSION
The dynein light chain, although a component of the cytoplasmic dynein motor complex, is promiscuous and has been identified as a component in numerous, diverse complexes. A universal role postulated for the dynein light chain is to serve as a dimerization machine. In S. cerevisiae, 25% of the Dyn2 cytoplasmic pool is found associated with the nuclear pore complex.   Fig. 1A (green, 100% identity; yellow, 80% identity, as determined across 12 diverse species), mapped on the Dyn2 dimer shown in surface representation. Nup159 pep2 is shown in stick format in purple, inserted in the highly conserved interdimer groove. The highly conserved QT substrate motif (green sticks) is located C-terminal to the Nup159_2 ␤-strand. Conservation, however, is equally distributed across the Dyn2 substrate binding region. The inset shows the relative orientation of the complex in graphic format colored as in Fig. 2A. B, electrostatic surface calculated using APBS to generate solventaccessible surface potentials that are shown in k B T/e, colored according to the key shown (42). Nup159 pep2 is shown in purple stick format with specific Dyn2 residues involved in hydrogen bond contacts labeled. The conserved QT motif is shown in green stick format. Dyn2-Nup159 pep2 interactions include Tyr-68 -Glu-1119, Glu-38Ј-Gln-1123, and Lys-12-Asp-1125. The complex is oriented as in A.
Nup159, a component of the Nup82 complex of the cytoplasmic fibrils, was identified as a Dyn2 binding partner that promotes stable association of the Nup82 complex with the NPC (11). The Nup159 pentameric array of Dyn2 binding sites links the N-terminal Phe-Gly repeat region with the C-terminal NPC anchor region (20).
Our analysis of Dyn2 homodimer binding to individual Nup159 peptides showed similar affinities for pep2 and pep4, at 17.9 and 13.1 M, respectively, whereas binding for pep1, pep3, and pep5 could not be experimentally determined based on properties of the individual peptides as synthesized. Binding curves fit best to a one-site binding model and are comparable with LC8 binding to peptides of similar size; DYNLL1 binds a 7-amino acid peptide from Bmf with a K D of 1.1 M and similarly sized neuronal NOS peptide with a K D of 7.0 M (36). The affinities determined between Dyn2 and the Nup159 peptides FIGURE 5. The Dyn2-Nup159 pep2 interaction is mediated by an extensive interaction network that recognizes 10 contiguous Nup159 residues, dually conferring specificity and substrate plasticity. A and B, close-up of residues involved in the Dyn2-Nup159 pep2 interaction. Secondary structure elements are shown as in Fig. 2A, with specific residues that mediate the Dyn2-Nup159 pep2 interaction shown in stick format and their corresponding hydrogen-bonding network shown with dashed lines. C, interaction matrix, showing the contact networks formed between Dyn2 protomers A and C with Nup159. Secondary structure elements and protomer designation are indicated along matrix axes. Backbone-backbone, backbone-side chain, side chain-side chain, and van der Waals interactions are indicated in blue, pink, red, and gray, respectively, and correlate with distances less than or equal to 3.5 (hydrogen bonds) and 4.5 Å (van der Waals contacts).
do not take into account potential cooperativity between arrayed Dyn2 homodimers based on interactions we observed in translational symmetry mates in the Dyn2-Nup159 crystal. The affinities and differential thermal binding modes determined for Nup159 pep2 and pep4 reflect plasticity in the Dyn2 binding pocket. The Dyn2 binding site does not have many steric occlusions and can thereby accommodate sequence diversity as observed with the LC8 family (25,36,37). The Dyn2-Nup159 crystal structure shows that extensive backbone-backbone interactions mediate the antiparallel ␤-sheet extension. This backbone-based interaction affords tight bind-ing while simultaneously enabling diversity in the side chains that flank the core, conserved QT binding motif. The QT motif is present in most Dyn2/LC8 binding peptides characterized to date and constitutes the C-terminal flank of the ␤-strand of the target peptide. The QT motif contributes a network of hydrogen bonds and van der Waals contacts with the conserved groove of the dynein light chain, directly contacting residues from each subunit of the homodimer. The amino acid diversity flanking the QT motif likely underlies the differential affinities and thermal binding modes observed across dynein light chain targets. Pep4 exhibited exothermic binding, indicative of a  Dyn2-Nup159 crystallographic symmetry mates are shown in graphic representation, colored as in Fig. 2A. Dyn2 interdimer interactions coupled with parallel, arrayed binding motifs on Nup159 likely promote linear, cooperative binding activity between Dyn2 dimers and Nup159. strong enthalpic, electrostatically driven interaction, whereas pep2 exhibited endothermic binding, indicative of a hydrophobic, entropically driven interaction. Pep2 and pep4 each have electrostatic and hydrophobic residues. A key hydrophobic determinant that may underlie the endothermic binding observed with Nup159 pep2 is the tyrosine residue at position 1117, 6 residues upstream of the QT motif (i.e. Gln-6; Fig. 1D). The corresponding residue in Nup159 pep4 is an alanine. The Nup159 pep2 Gln-6 tyrosine makes numerous van der Waals contacts with the Dyn2 homodimer (Fig. 5C). Peptide-specific exothermic and endothermic binding has been observed with Dyn2 homologs from other species and highlights the sequence diversity within target sites that dynein light chains are capable of accommodating (27, 36 -40).
Our work represents the first biophysical and structural characterization of the yeast dynein light chain, Dyn2. At physiological conditions, Dyn2 exists predominantly in the homodimeric state. As a homodimer, Dyn2 is positioned to interact with target sites and induce and stabilize parallel dimerization in these target proteins. Dimerization machines can cross-link targets, homo-or heterodimerize targets, serve to architecturally extend a target, as well as promote the avidity of a target for binding partners. The Dyn2-Nup159 structure creates a foundation for understanding the role of Dyn2 in the NPC as a dimerization machine that can scaffold Nup159 and extend the protein at least 170 Å (5 ϫ 34 Å). Our structural and biophysical investigations of the Dyn2-Nup159 interaction have additional implications for the Dyn2 mode of interaction and function with the dynein intermediate chain, Pac11, and for its potential role in promoting Pac11 dimerization and aiding in the recruitment of the dynein activation complex, dynactin (21).