Probing the Potential Glycoprotein Binding Site of Sindbis Virus Capsid Protein With Dioxane and Model Building

Transcription

1 PROTEINS: Structure, Function, and Genetics 33: (1998) Probing the Potential Glycoprotein Binding Site of Sindbis Virus Capsid Protein With Dioxane and Model Building Sukyeong Lee, Richard J. Kuhn, and Michael G. Rossmann* Department of Biological Sciences, Purdue University, West Lafayette, Indiana ABSTRACT Alphavirus budding from the plasma membrane is initiated by the specific interaction of the nucleocapsid with the cytoplasmic domain of the glycoprotein E2. It was proposed (Lee et al., Structure 4: , 1996) that binding of the capsid protein residues 108 to 110 (the N-terminal arm residues) to a hydrophobic pocket on the surface of the neighboring capsid protein in the crystal structure mimics the binding of the E2 C-terminal residues into this pocket. In addition, structural comparisons of wild-type and mutant Sindbis virus capsid protein (SCP) and Semliki Forest virus capsid protein suggested that budding is associated with a switch between two conformations of the hydrophobic pocket. To test the proposed mechanism, SCP( ), which is missing the N-terminal arm, was crystallized to examine the pocket conformation when the pocket is empty. However, the pocket was occupied by dioxane molecules from the crystallization solution. The pocket conformation was the same as that when it was occupied by the N-terminal arm, demonstrating that the pocket favors binding ligands of appropriate size and shape. Proteins 33: , Wiley-Liss, Inc. Key words: alphavirus; capsid structure; budding; dioxane; assembly INTRODUCTION Alphaviruses are simple enveloped viruses with a single positive RNA genome. Both the nucleocapsid and the glycoprotein spikes are arranged with quasi T 4 icosahedral symmetry. Each spike consists of trimers of heterodimers of E1 and E2 glycoproteins which are anchored in the viral membrane by their membrane-spanning C-terminal region. E1 and E2 have 2- and 33-residue cytoplasmic domains, respectively. The E2 cytoplasmic domain interacts with the nucleocapsid 1 and is essential for virus budding, but the E1 cytoplasmic domain is dispensable. 2 Molecular genetic and biochemical studies have shown that several residues which are involved in palmitylation 3 or phosphorylation 4,5 of the E2 cytoplasmic domain are important for budding. Studies with chimeric viruses of Sindbis (SINV) and Ross River (RRV) viruses suggested that Ala403 and Asn405 in the E2 cytoplasmic domain are also important for budding. 6 Collier et al. have shown that peptides corresponding to sequences of the E2 cytoplasmic domain of SINV can inhibit virus budding. 7 They hypothesized that these peptides competitively inhibit attachment of E2 to the nucleocapsid. These peptides include Tyr400, subsequently also identified by mutagenesis as important for budding. 8,9 Only a few studies have been conducted to identify the regions in the capsid protein which are involved in the binding of the glycoprotein E2. Molecular genetic and chemical cross-linking studies 10,11 have shown that Tyr180 or Glu183 in SCP were involved in E2-capsid interaction. We previously proposed, based on structural studies, 1 that the major interactions between the E2 cytoplasmic domain and alphavirus cores were centered at a hydrophobic pocket on the capsid protein. In a variety of wild-type and mutant SCP structures, the pocket was occupied by Leu108 and Leu110 in the N-terminal arm from neighboring SCP molecules. Sequence analogy and structural studies suggested that the N-terminal arm residues Leu108 and Leu110 were analogous to residues Tyr400 and Leu402 of the E2 glycoprotein. 1 Genetic studies 12 supported this model. In contrast to SCP, the pocket was empty in structures of Semliki Forest virus capsid protein (SFCP) where Tyr180 and Trp247 had large conformational changes. 13 This suggested a mechanism in which the structure of the capsid protein was switched to a different conformation on binding E2. 1 Alternatively it is possible that the structural differences between SCP and SFCP are intrinsic properties of the two viruses and, therefore, unrelated to the N-terminal arm binding. To resolve this problem it is necessary to compare the structures of the same virus capsid protein in its The numbering is according to SINV. Grant sponsor: National Institutes of Health; Grant numbers: AI33982, AI11219, and AI *Correspondence to: Michael G. Rossmann, Department of Biological Sciences, Purdue University, West Lafayette, IN mgr@indiana.bio.purdue.edu Received 3 March 1998; Accepted 8 May WILEY-LISS, INC.

2 312 S. LEE ET AL. TABLE I. X-ray Diffraction Data Number of measurements [I 1 (I)] Number of unique reflections [I 1 (I)] Data set Overall SCP R sym* (%) ( ) % observed Overall SCP R sym (%) ( ) % observed peptide (399-PYALP- 404) soaked *R sym (%) ( h i 0 I hi 7I h 80)/( h i I hi ) 100, where I hi is an individual observation and 7I h 8 is the mean intensity for reflection h. bound (to the N-terminal arm) and unbound form. Therefore, SCP( ), which is missing the N-terminal arm, was crystallized to examine the conformation of the empty pocket. Contrary to expectation, the hydrophobic pocket was found to have a conformation similar to the pocket with bound N-terminal arm. However, dioxane molecules, which had been used for crystallization, occupied the pocket in a similar way to that of the two leucine side chains in the N-terminal arm bound structure. MATERIALS AND METHODS Crystallization SCP( ) was expressed in E. coli and purified using the same strategy previously described. 14 The hanging drop vapor diffusion method was used at room temperature for crystallizing SCP( ). Each drop was made up of 3.5 µl of protein solution (21 mg/ml protein in 50 mm Tris buffer, ph 7.6; 25 mm sodium chloride; and 1 mm EDTA) and 3.5 µl of reservoir solution. Reservoir solutions contained 20 to 26% (w/v) PEG 8000; 100 mm sodium cacodylate buffer, ph 6.5; 150 mm sodium acetate; and 6% (v/v) dioxane. Microseeding was necessary to grow the crystals bigger, and dioxane was required to obtain well ordered single crystals. Data Collection The crystals of SCP( ) were flash-frozen on a loop in a nitrogen stream at 110 K. Before freezing, crystals were soaked for an hour in cryo-solvent consisting of 25% (v/v) glycerol; 22% (w/v) PEG 8000; 100 mm sodium cacodylate buffer, ph 6.5; 150 mm sodium acetate; and 4% (v/v) dioxane. A diffraction data set of SCP( ) was collected on an R-axis IIc detector using a CuK rotating anode X-ray source (Rigaku RU-H2R) operated at 50 kv and 100 ma with bent focusing mirrors. The crystal-todetector distance was 11 cm. An oscillation angle of 1.0 was used for the first 38 of data, and each image was exposed for 10 minutes. An oscillation range of 1.3 and an exposure time of 20 minutes were used for the next 165 of data. The crystal diffracted to 2.0 Å resolution. The HKL package 15 was used to process the data. The space group was found to be P1 with unit cell dimensions a 36.0 Å, b 59.5 Å, c 71.1 Å, 109.4, 101.5, Data collection statistics are shown in Table I. The redundancy of the unique reflections was 1.9. The R sym of the total data set was 2.8% and the R sym for the highest resolution shell was 7.8%. The completeness of the data set was 78.3%. Crystals in 7 µl drops were soaked in various concentrations of the E2 hexapeptide (399-PYALAP- 404) by adding 10 µl of the peptide solution in mother liquor. The final concentration of the peptide was at least 10 mg/ml. After 4 days of peptide soaking, all the crystals had cracked. After 9 hours of soaking, a crystal was plunged into the same cryosolvent used for the unsoaked crystals. The crystal diffracted to 3.0 Å resolution with the same unit cell dimensions as the unsoaked crystal. However, the mosaicity was very high, 2.3 compared to 0.6 of the unsoaked crystal. A diffraction data set was collected in a similar way as had been used for the unsoaked crystals (Table I). Structure Determination A cross rotation function (program AMoRe 16 )was calculated using a monomer model of SCP derived from the SCP(S215A, ) crystal structure. 1,14 Typical results were obtained using data between 10.0 and 3.0 Å resolution and a radius of integration of 25 Å. Four peaks were found with heights of about 20 arbitrary units, while all other peaks were less than 10 units. The top four solutions were reproducible, independent of the resolution of the data used in the calculations. Since the space group is P1, the center of gravity of one molecule was arbitrarily positioned at the origin. A translation function search (program AMoRe) was then calculated for each of the other three molecules (molecules 2, 3, and 4) relative to the reference molecule, using data between 8.0 and 3.5 Å resolution. In each case, a distinct solution was found. Structure factors calculated by placing

3 GLYCOPROTEIN BINDING TO SINDBIS NUCLEOCAPSID 313 Fig. 1. SCP( ) crystals have a P1 unit cell (dashed outline) with a pseudo face-centered monoclinic packing arrangement (solid outline) in which the two-fold axis is along the a axis of the P1 cell (a 36.0 Å, b 59.5 Å, c 71.1 Å, 109.4, 101.5, and 90.1 ). The unit cell dimensions of the pseudo B2 (a axis unique) space group derived from the P1 unit cell dimensions are a 36.0 Å, b 59.5 Å, c Å, 109.4, 93.2, and In this pseudo B2 space group packing arrangement, the two molecules in the pseudo asymmetric unit form a dimer related by 177. The dimer axis (shown as a thick arrow) is tilted by about 20 to the pseudo crystallographic two-fold axis. The origin of the triclinic cell is chosen arbitrarily at the center of molecule 1. Molecules at level 1/2 are indicated by unfilled circles. the four independent molecules into the unit cell had a correlation coefficient (CC) of and an R-factor of 34.2% with respect to the observed amplitudes. The translation function search was followed by a rigid body refinement of the rotation angles and positions of each molecule using the program AMoRe. SCP( ) crystals have a P1 unit cell with a pseudo face-centered monoclinic packing arrangement in which the two-fold axis is along the a axis of the P1 cell (Fig. 1). In this pseudo B2 space group arrangement, two molecules are in the pseudo asymmetric unit forming a dimer with monomer-monomer contacts similar to that found in two other crystal forms of SCP. 14 The two molecules (1 and 2 or 3 and 4) forming the dimer in the SCP( ) structure are related by 177 rather than a 180 rotation, and the dimer axis is tilted by about 20 to the pseudo crystallographic two-fold axis. Refinement The structure was refined with the program X- PLOR using data between 8.0 and 2.0 Å resolution. Five percent of the data, represented by 1365 randomly selected reflections, was excluded from the working set and used to calculate R free. 18,19 The structure was refined, without any initial model building, using conjugate gradient minimization and one round of simulated annealing with a strong non-crystallographic symmetry (NCS) restraint among the four molecules in the unit cell. The model for each molecule was then rebuilt using omit maps anda(2fo-fc) map. Individual B-factors were applied with a strong NCS restraint. Density averaging combined with solvent flattening and histogram matching was applied using the program DM 20 starting with the phases obtained from the refined coordinates of the four molecules. These phases were TABLE II. Crystallographic refinement of SCP( ) Resolution of included data (Å) [I 1 (I)] R working (%) (number of reflections) 19.7 (26881) R free (%) (number of reflections) 26.8 (1378) r.m.s. deviation of bond lengths from standard (Å) r.m.s. deviation of bond angles from standard ( ) 1.7 r.m.s. deviation of improper angles from standard ( ) 1.6 r.m.s. deviation of dihedral angles from standard ( ) 26.9 Average B (main-chain) (Å 2 ) 13.8 Average B (side-chain) (Å 2 ) 17.2 Number of atoms without waters 4726 Number of waters 469 Number of reflections per refineable parameter 1.3 extended gradually from 3.0 to 2.0 Å resolution in 100 iterations. The averaged map was used to rebuild two flexible surface loops from residues 120 to 124 and from 170 to 176. Other model changes included the addition of the N-terminal methionine, the product of the initiation codon for protein translation. 82 NCS-related water molecules were modeled based on an averaged map, a (2Fo-Fc) map, and an (Fo-Fc) map. The correlations between the electron densities of molecules 2 and 3 and between molecules 1 and 4 were higher than other pairwise comparisons. Therefore, the restraint imposed for the four molecules was replaced by two pairwise restraints. More water molecules were added after further model building and 4 consecutive refinement cycles. At this stage, one dioxane molecule was

4 314 S. LEE ET AL. Fig. 2. Stereo image of the 4 molecules in SCP( ) crystals superimposed on the averaged electron density map (gray). Molecules 1 and 4 are shown in magenta. Molecules 2 and 3 are shown in blue. Since strong NCS restraints were applied in a pairwise manner between molecules 1 and 4 and between molecules 2 and 3 during refinement, only two different conformations can be seen in this figure. The dioxane molecules were excluded in the NCS restraints and, hence, four different dioxane structures are shown in the first half of the pocket. However, dioxane was modeled only for molecule 2 in the second half of the pocket. Fig. 3. The hydrophobic pocket of SCP( ) in molecule 2 with the N-terminal arm superimposed. The models of dioxane and of the residues in the hydrophobic pocket are colored. The magenta color is the N-terminal arm superimposed from the SCP(S215A, ) crystal structure. The electron density map isa(2fo-fc) map with phases calculated from the model shown here. The map was calculated using data between 15 and 2 Å resolution. The electron density of the pocket is shown as salmon, and the electron density of the two dioxane molecules is shown in white. The positions of the side chain of two leucines in the arm match with the positions of dioxanes.

5 GLYCOPROTEIN BINDING TO SINDBIS NUCLEOCAPSID 315 Fig. 4. Compensational sequence changes in the N-terminal arm and the hydrophobic pocket of alphaviruses. The altered residues are shown with the same color in the diagrammatic figure (top panel) and table (bottom panel). For instance, a phenylalanine at position 110 (green) is associated with an isoleucine (green) at position 137 in AURV. Abbreviations: SINV, Sindbis virus; AURV, Aura virus; WEEV, Western equine encephalitis virus; EEEV, Eastern equine encephalitis virus; VEEV, Venezuelan equine encephalitis virus; ONNV, O nyong-nyong virus; RRV, Ross River virus; SFV, Semliki Forest virus. introduced into the hydrophobic pocket of molecule 2 and an N-formyl group was added to each N-terminus. After some further cycles of refinement, dioxane molecules were modeled into the other three protein molecules, and a second dioxane molecule was modeled into the hydrophobic pocket of molecule 2. The final refinement cycles of the atomic coordinates used conjugate gradient minimization with no NCS restraints among dioxane molecules, but with pairwise restraints among protein molecules. The final model included 4 molecules of SCP( ), four N-formyl methionines at the N-termini, 5 molecules of dioxane, and 469 molecules of water. The final R working factor was 19.7%, R free was 26.8%, and the r.m.s. deviation of bond lengths from ideal geometry 21 was Å. A Ramachandran plot showed that 89.0% of the residues were in the most favored region, and 11.0% of the residues were in the additionally allowed region. Detailed statistics of refinement are shown in Table II.Coordinates have been deposited with the Brookhaven Protein Data Bank with code 1wyk.

6 316 S. LEE ET AL. RESULTS AND DISCUSSION Conformation of the Hydrophobic Pocket The truncated SCP( ) was generated to determine the structure of the empty SCP hydrophobic pocket and to make crystals of SCP complexed with E2 peptides. Difference maps between the native and peptide (E2 glycoprotein residues, 399- PYALAP-404) soaked crystals showed no evidence of bound peptide. However, these crystals do not have much space either to accommodate the hexapeptide or to allow diffusion of the peptide into the pocket. If the hexapeptide were bound to the pocket of molecules 1 or 4 in a manner analogous to the N- terminal arm, then SCP residues Arg207 and Thr238 from a neighboring unit cell would clash with Pro399 of the E2 peptide. Similarly, if the peptide were bound to molecules 2 or 3, Pro399 would clash with Arg212, and Ala403 would clash with Glu263, Leu140, and His141 of a neighboring molecule. According to the proposed budding mechanism, 1 the capsid structure without the bound E2 cytoplasmic residues would have Tyr180 and Trp247 as found in SFCP, where Tyr180 is pointing out of the pocket and Trp247 is placed close to the space occupied by Tyr180 in SCP. However, the hydrophobic pocket of all four SCP( ) molecules described here has the conformation of Tyr180 and Trp247 as if the N-terminal arm were bound into the pocket. Although there is no N-terminal arm available to bind into the pocket, there are one or two dioxane molecules in the pockets of each of the four SCP( ) molecules (Fig. 2), situated as are Leu108 and Leu110 when the N-terminal arm binds into the pocket. The hydrophobic pocket can be divided into two compartments with Trp247 in the middle of the pocket. The first half of the pocket, lined by Tyr180 and Trp247, is occupied by Leu108 and the second half of the pocket, lined by Trp247 and Phe166, is occupied by Leu110 when the N-terminal arm is present. The first part of the pocket is smaller than the second half of the pocket. In the SCP( ) structure, the first half of the hydrophobic pocket is occupied by a dioxane molecule in each of the four independent molecules (Fig. 2). The dioxane had been added to grow well-shaped single crystals. The position of dioxane in the pocket matches well with the position otherwise occupied by Leu108 which is analogous to Tyr400 in the E2 cytoplasmic domain (Fig. 3). Dioxane molecules were modeled independently for each of the four molecules. As there are no polar interactions that might provide specific orientation to the liganded molecules, the dioxane models represent the probable variation in orienting these molecules in the pocket. The second half of the hydrophobic pocket has some extra density in each molecule, but it is not as definitive as in the first half of the pocket. For molecule 2, a dioxane molecule was modeled into the second half of the pocket (Fig. 3), but water molecules were modeled into the second half of the pocket for the other three SCP molecules. There may be some exchange of dioxane with these water molecules. The larger, second half of the pocket does not bind dioxane as well as the first half of the pocket, because in that pocket, the dioxane has fewer hydrophobic contacts. The results given here do not determine whether the pocket would have a structure similar to the conformation of that found in SFCP structures if there were no dioxane in the pocket. However, they do show the pocket has an affinity for suitable ligands. It would seem unlikely that such a property is accidental, but rather represents a functionally important property. We had suggested 1 that the presence of ligands in the pocket would stabilize the core structure as required in the formation of an infectious particle. Similarly, the removal of the glycoprotein spikes during cell entry would destabilize cores in readiness for disassembly. Interaction of the Hydrophobic Pocket With the N-Terminal Arm and With E2 in Different Members of Alphaviruses The two leucine residues in the N-terminal arm are important for capsid assembly in the cell. 1 However, in contrast to the equivalent E2 glycoprotein residues Tyr400 and Leu402, which are completely conserved among alphaviruses, the N-terminal arm residues 108 and 110 have variability although their hydrophobicity is conserved (Fig. 4). Alphaviruses, which have a methionine in the 108 position, have a valine at position 132 and glycine at 133 (Fig. 4). When methionine was modeled in place of Leu108, there was steric hindrance with the side chain of Met132 and Glu133 in SCP. However, the residues 132 and 133 are valine and glycine when residue 108 is methionine, which relieves the steric interference, showing that there are complementary changes in these viruses relative to Sindbis virus. Leu110 in SCP corresponds to an isoleucine in some alphaviruses, including SFV. The isoleucine side chain would interfere with Phe166 in SCP. This is avoided in these viruses by having a cysteine in place of a phenylalanine (Fig. 4). In Aura virus (AURV), Leu110 is changed to phenylalanine. The phenylalanine side chain would interfere with Met137 or Met164 in SCP. However, Met137 is changed to isoleucine in AURV, which relieves the steric hindrance with the phenylalanine. In all modeling exercises, care was taken to avoid steric interference of the modeled side chains with the conserved Tyr180 and Trp247. The model building showed that changes of amino acids in the N-terminal arm compensate the sequence changes in the hydrophobic pocket.

7 GLYCOPROTEIN BINDING TO SINDBIS NUCLEOCAPSID 317 The Binding of the Glycoprotein to Capsid During Budding Sequence differences among alphaviruses suggest that the pocket and N-terminal arm undergo compensational changes. This implies that the interactions between the N-terminal arm and pocket are specific and optimized. However, residues Tyr400 and Leu402 in the cytoplasmic domain of E2 (analogous to Leu108 and Leu110 in the N-terminal arm) are completely conserved, showing that E2 has not adapted itself to the pocket as well as has the N-terminal arm. Possibly the binding of E2 into the hydrophobic pocket is aided by association of the palmitylated cysteines (416 and 417) with the membrane. Mutagenesis studies showed the importance of the hydroxyl group of Tyr400. 8,9,22 The most probable candidate for hydrogen bonding with the phenolic group of Tyr400 is the main chain carbonyl group of capsid residue 132. If this interaction does occur, it would explain why tyrosine is highly conserved whereas residues lining the pocket are more variable. These observations are consistent with the hypothesis that the cytoplasmic domain of E2 displaces the N-terminal arm of the capsid protein during the assembly process. ACKNOWLEDGMENTS We thank Zhu Zhu and Bonnie McKinney for excellent technical assistance and Katherine E. Owen for stimulating discussions. The work was supported by NIH grants to RJK (AI33982) and MGR (AI11219), an NIH program project grant (AI35212), a Lucille P. Markey grant in support of structural biology at Purdue University and a Purdue University reinvestment grant. REFERENCES 1. Lee, S., Owen, K.E., Choi, H.-K. et al. Identification of a protein binding site on the surface of the alphavirus nucleocapsid and its implications in virus assembly. Structure 4: , Barth, B.U., Suomalainen, M., Liljeström, P., Garoff, H. Alphavirus assembly and entry: Role of the cytoplasmic tail of the E1 spike subunit. J. Virol. 66: , Ivanova, L., Schlesinger, M.J. Site-directed mutations in the Sindbis virus E2 glycoprotein identify palmitoylation sites and affect virus budding. J. Virol. 67: , Liu, N., Brown, D.T. Phosphorylation and dephosphorylation events play critical roles in Sindbis virus maturation. Virology 196: , Liu, L.N., Lee, H., Hernandez, R., Brown, D.T. Mutations in the endo domain of Sindbis virus glycoprotein E2 block phosphorylation, reorientation of the endo domain, and nucleocapsid budding. Virology 222: , Lopez, S., Yao, J.-S., Kuhn, R.J., Strauss, E.G., Strauss, J.H. Nucleocapsid-glycoprotein interactions required for assembly of alphaviruses. J. Virol. 68: , Collier, N.C., Adams, S.P., Weingarten, H., Schlesinger, M.J. Inhibition of enveloped RNA virus formation by peptides corresponding to glycoprotein sequences. Antivir. Chem. Chemother. 3:31 36, Gaedigk-Nitschko, K., Schlesinger, M.J. Site-directed mutations in Sindbis virus E2 glycoprotein s cytoplasmic domain and the 6K protein lead to similar defects in virus assembly and budding. Virology 183: , Zhao, H., Lindqvist, B., Garoff, H., von Bonsdorf, C.H., Liljeström, P. A tyrosine-based motif in the cytoplasmic domain of the alphavirus envelope protein is essential for budding. EMBO J. 13: , Lee, H., Brown, D.T. Mutations in an exposed domain of Sindbis virus capsid protein result in the production of noninfectious virions and morphological variants. Virology 202: , Lee, H., Ricker, P.D., Brown, D.T. The configuration of Sindbis virus envelope proteins is stabilized by the nucleocapsid protein. Virology 204: , Owen, K.E., Kuhn, R.J. Alphavirus budding is dependent on the interaction between the nucleocapsid and hydrophobic amino acids on the cytoplasmic domain of the E2 envelope glycoprotein. Virology 230: , Choi, H.-K., Lu, G., Lee, S., Wengler, G., Rossmann, M. G. Structure of Semliki Forest virus core protein. Proteins 27: , Choi, H.-K., Lee, S., Zhang, Y.-P. et al. Structural analysis of Sindbis virus capsid mutants involving assembly and catalysis. J. Mol. Biol. 262: , Otwinowski, Z., Minor, W. Processing of X-ray diffraction data collected in oscillation mode. In: Methods in Enzymology. Carter C.W., Jr., and Sweet, R.M. (eds.). Vol New York: Academic Press. 1997: Navaza, J. AMoRe: an automated package for molecular replacement. Acta Crystallogr. A50: , Brünger, A.T. X-PLOR. Version 3.1. A System for X-ray Crystallography and NMR. New Haven and London: Yale University Press, Brünger, A.T. The Free R value: a novel statistical quantity for assessing the accuracy of crystal structures. Nature 355: , Brünger, A.T. Assessment of phase accuracy by cross validation: The free R value. Methods and applications. Acta Crystallogr. D49:24 36, Collaborative Computational Project Number 4. The CCP4 suite: Programs for protein crystallography. Acta Crystallogr. D50:60 763, Engh, R.A., Huber, R. Accurate bond and angle parameters for X-ray protein structure refinement. Acta Crystallogr. A47: , Skoging, U., Vihinen, M., Nilsson, L., Liljeström, P. Aromatic interactions define the binding of the alphavirus spike to its nucleocapsid. Structure 4: , 1996.