Glycoprotein Organization of Chikungunya Virus Particles (2010) PDF

LETTER doi:10.1038/nature09555 Glycoprotein organization of Chikungunya virus particles revealed by X-ray crystallography James E. Voss1,2, Ma...

LETTER doi:10.1038/nature09555 Glycoprotein organization of Chikungunya virus particles revealed by X-ray crystallography James E. Voss1,2, Marie-Christine Vaney1,2, Stéphane Duquerroy1,2,3, Clemens Vonrhein4, Christine Girard-Blanc5,6, Elodie Crublet5,6, Andrew Thompson7, Gérard Bricogne4 & Félix A. Rey1,2 Chikungunya virus (CHIKV) is an emerging mosquito-borne alpha- make a glycoprotein shell enclosing the viral membrane and the virus that has caused widespread outbreaks of debilitating human nucleocapsid. E3, which contains the 64 amino-terminal residues of disease in the past five years1. CHIKV invasion of susceptible cells is p6213, remains peripherally attached to virions for some alphaviruses14. mediated by two viral glycoproteins, E1 and E2, which carry the Alphaviruses enter cells via receptor-mediated endocytosis15. The main antigenic determinants and form an icosahedral shell at the acidic endosomal environment triggers an irreversible conformational virion surface. Glycoprotein E2, derived from furin cleavage of the rearrangement of the surface glycoprotein shell of mature virions16. p62 precursor into E3 and E2, is responsible for receptor binding, The E2–E1 heterodimer dissociates and E1 rearranges into fusogenic and E1 for membrane fusion. In the context of a concerted multi- homotrimers that induce fusion of viral and endosomal mem- disciplinary effort to understand the biology of CHIKV2, here we branes17,18, allowing the release of the viral nucleocapsid into the cyto- report the crystal structures of the precursor p62–E1 heterodimer sol. The crystal structure of the alphavirus E1 ectodomain has been and of the mature E3–E2–E1 glycoprotein complexes. The resulting determined in both pre-19 and post-fusion20 conformations, showing atomic models allow the synthesis of a wealth of genetic, bio- that it is folded into three b-sheet-rich domains (domains I, II and III; chemical, immunological and electron microscopy data accumu- Fig. 1a), with an internal fusion loop at the tip of domain II21. The lated over the years on alphaviruses in general. This combination structure of alphavirus E2, in contrast, has remained elusive over the yields a detailed picture of the functional architecture of the 25 MDa years. We adopted the strategy described in Methods to produce, crys- alphavirus surface glycoprotein shell. Together with the accompany- tallize and determine the 2.2-Å-resolution structure of recombinant ing report on the structure of the Sindbis virus E2–E1 heterodimer at CHIKV p62–E1 and furin-processed, mature E3–E2–E1 glycoprotein acidic pH (ref. 3), this work also provides new insight into the acid- complexes. triggered conformational change on the virus particle and its inbuilt The structure shows that the p62–E1 heterodimer has the shape of a inhibition mechanism in the immature complex. twisted plate about 150 Å long, 50 Å wide and 25 Å thick (Fig. 1a and Although the symptoms of the disease were first described in the Supplementary Fig. 3), with E3 sticking out at one side (Fig. 1a inset, eighteenth century4, CHIKV was first isolated only in 1952 during a and Supplementary Fig. 3a). The furin site is in an exposed loop at the dengue outbreak in Tanzania5. The name Chikungunya means surface of the spike (Fig. 1a). The mature E3–E2–E1 complex is very ‘‘stooped walk’’ in the Tanzanian Kimakonde language6, an allusion similar, the only important difference being in the residues forming the to the persistent arthralgia caused by the disease, which otherwise shares furin loop, which become disordered on cleavage. E3 is an a/b protein, clinical similarities with dengue disease7. CHIKV is geographically with an N-terminal b-hairpin packing against three a-helices orga- spread throughout vast regions of Africa and Asia, but remained essen- nized into a horseshoe shape (Supplementary Fig. 4c). E2 is an all b tially neglected until an important epidemic outbreak in 2005 in islands protein belonging to the immunoglobulin superfamily, with three of the Indian Ocean8 attracted the attention of the western world. This immunoglobulin domains labelled A, B and C in amino- to carboxy- outbreak was the consequence of an adaptation of CHIKV to efficiently terminal order (Fig. 1b). Domain B is at the membrane distal end and infect Aedes albopictus mosquitoes in addition to the normal vector domain C is towards the viral membrane, with domain A at the centre. Aedes aegypti, and important determinants of the new vector adaptation Domain B is presented at the tip of a long b-ribbon connector (Sup- were mapped to the CHIKV envelope proteins9,10. The presence of the plementary Fig. 4b), which makes most of the contacts with E3 (Sup- new mosquito vector in many areas of Europe and the Americas11 raises plementary Fig. 5). A detailed description of the individual domains is concerns of further expansion of the endemic zones. provided in Supplementary Information. p62 and E1 are type I membrane proteins that are derived from a In the complex, E1 makes no direct contact with E3, interacting structural polyprotein precursor (Supplementary Fig. 1a). Their laterally with E2 all along domain II. Furthermore, the segment imme- cotranslational association, in the endoplasmic reticulum of the diately downstream of domain III makes an additional strand inserted infected cell, into a p62–E1 heterodimer is required for proper folding. in E2 domain C. E1 is bent compared to its structure in isolation, with In turn, the heterodimers trimerize to form the viral ‘spikes’. Furin the fusion loop unfolded to make a short b-hairpin inserted in a groove maturation of p62 into E3 and E2 during transport to the cell surface between domains A and B (Figs 1a and 2). This conformation is primes the spikes for subsequent fusogenic activation for cell entry. stabilized by a number of E2 histidine side chains (Fig. 2a), some of Mature virions bud at the plasma membrane via interactions between which hydrogen bond to the main chain of the E1 fusion loop. The area E2 and genome-containing viral nucleocapsids present in the cyto- buried from solvent per protomer in the E2–E1 complex is about 2,500 plasm. The recently reported cryo-electron microscopy (cryo-EM) Å2 (Supplementary Fig. 3 and Supplementary Table 4a), which is large reconstruction of CHIKV virion-like particles confirmed that all alpha- compared to the 1,100 Å2 buried in the flavivirus E homodimer22. The viruses have a common architecture12. The particles are organized with a-helical portion of E3 packs against the E2 b-ribbon, burying 800 Å2 icosahedral symmetry of triangulation T 5 4, containing 80 spikes that of its surface from solvent (Supplementary Table 4b) at the region 1 Institut Pasteur, Département de Virologie, Unité de Virologie Structurale, 25 rue du Dr Roux, 75724 Paris Cedex 15, France. 2CNRS URA 3015, 25 rue du Dr Roux, 75724 Paris Cedex 15, France. 3 Université Paris-Sud, Faculté d’Orsay, 91405 Orsay Cedex, France. 4Global Phasing Ltd, Sheraton House, Castle Park, Cambridge CB3 0AX, United Kingdom. 5Institut Pasteur, Département de Biologie Structurale et Chimie, Plateforme de Production de protéines recombinantes, 25 rue du Dr Roux, 75724 Paris Cedex 15, France. 6CNRS URA 2185, 25 rue du Dr Roux, 75724 Paris Cedex 15, France. 7 Synchrotron SOLEIL, L’Orme de Merisiers, BP 48 St Aubin, 91192 Gif sur Yvette, France. 2 D E C E M B E R 2 0 1 0 | VO L 4 6 8 | N AT U R E | 7 0 9 ©2010 Macmillan Publishers Limited. All rights reserved  RESEARCH LETTER a b E3 N terminal N terminal 50 Å II I N terminal Furin loop Furin loop B A C III 1 E2 domain A N12 E2 β-ribbon E3 2 Central E3 E1 domain III E2 domain B N263 E2 domain C 3 arch N263 N141 C terminal 7 8 6 Arch 1 N12 E1 FL 9 Arch 2 4 E1 domain I 5 E2 domain B E2 domain A E2 domain C E1 domain II Membrane Figure 1 | Structure of the p62–E1 heterodimer. a, Ribbon diagram of the arrow (next to the pink and blue stars indicating the C termini of p62 and E1, p62–E1 heterodimer. E1 domains I, II and III are shown in red, yellow and blue, respectively) points to the viral membrane. Inset, schematic diagram, with the respectively, and the fusion loop (FL) in orange. E3 is coloured white/grey and heterodimer ‘plate’ drawn ‘untwisted’, showing how the domains are E2 domain A is coloured cyan, B dark green, C pink and the b-ribbon dark positioned with respect to one another and their connectivity. b, p62 purple. The N-linked glycans are shown in ball and stick, coloured according to organization, oriented roughly at 90 degrees from a to show E3. Green numbers atom type, and labelled. The disulphides are depicted as green sticks. The black label the disulphides according to Supplementary Fig. 1b. behind the groove in which the E1 fusion loop inserts. The structure We built a model for the whole glycoprotein layer of the alphavirus shows that E3 acts as a brace, maintaining domain B in an orientation particle (Fig. 3) by fitting the structure of the CHIKV E2–E1 hetero- with respect to domain A such that it creates the groove accommodating dimer into the available cryo-EM reconstructions of alphavirus particles the fusion loop. This is consistent with the observation that recombinant as explained in Supplementary Information (Supplementary Fig. 6). E2 lacking the E3 moiety does not dimerize with E123. The result gives the intra- and interspike contact details (Supplemen- tary Table 4c, d), confirming that E2 and E1 make all the intra- and interspike contacts, respectively. Domain A makes three-fold contacts at a the spike top, with domain B projecting to the side to give the spike its characteristic propeller shape. Domain C is sandwiched between domain II of two neighbouring E1 molecules, making the walls of a E2 domain A central cavity under domain A, centred on the three-fold axis of the spike. E2 domain B The inferred atomic model is consistent with the available biological H73 W89 data on alphaviruses, with mutations conferring escape to neutralizing 5 antibodies mostly clustered on domain B, as well as the top of domain A (Supplementary Figs 1b, 7 and Supplementary Table 6). Furthermore, H29 4 the mutations affecting tissue tropism and host range also map to H18 8 A92 domains A and B, as described in Supplementary Information and H226 F87 9 Y93 illustrated in Supplementary Fig. 1b and Supplementary Table 7. In K202 N200 S57 particular, a number of determinants for efficient mosquito midgut F95 A226 2 Y85 infectivity of CHIKV24 and other alphaviruses (Supplementary Table 7c) map to the ‘wings’ and the FG loop (see Supplementary Informa- 3 tion) at the top of domain A. It is noteworthy that domain A, exposed at E1 fusion loop the top of the spike, is clearly homologous to domain III of the flavivirus E1 domain II envelope protein E (ref. 22) (Supplementary Fig. 4a and Supplementary 4 Figure 2 | The E1 fusion loop binding groove in E2. a, Fusion loop contacts. Domain III E2 (in a transparent surface rendering) and E1 (in ribbons) are in the b background and foreground, respectively. Dotted white lines show hydrogen Fusion loop bonds from E2 histidines to the fusion loop main chain. Residues marking the ij loop Domain II E2 transitional epitope26,28, are coloured red and labelled. Green numbers highlight the cluster of disulphide bonds in this part of the complex. The residue Ala 226 is shown as a grey sphere. b, E1 conformation. The 3 Å Semliki Forest virus (SFV) E1 model (PDB code 2ALA), coloured by domains with the fusion loop cyan, is shown superposed on E1 from the p62–E1 complex, in grey with the fusion loop orange. The black arrow emphasizes the movement of the 50 Å Domain I fusion loop and the adjacent ij loop, which contains residue Ala 226, mentioned in the text. 7 1 0 | N AT U R E | VO L 4 6 8 | 2 D E C E M B E R 2 0 1 0 ©2010 Macmillan Publishers Limited. All rights reserved  LETTER RESEARCH Table 2). A small patch of sequence similarity had indeed been a c SINV N196 detected25, corresponding to the connection between two homologous glycan b-strands (the EF loop) in the corresponding immunoglobulin b-barrels. 200 202 Virulence determinants in flavivirus E protein also map to the FG loop of 2 2 3 domain III, which has a similar orientation in the flavivirus particle q3 q3 (Supplementary Fig. 4a). The functional data provide further support 5 2 5 381 Ct to propose a common ancestry for these domains, which are inserted 300 381 Stem 361 361 300 within proteins that are otherwise not homologous, confirming the mosaic nature of genes coding for viral envelope proteins in general. Viral membrane TM Particularly informative are the locations of transitional epitopes, identified in E2 (residues 200 and 202), and in E1 (residues 300, 361 20 nm Capsid and 381)26, which were reported to become accessible only on exposure b d of alphavirus particles to certain treatments, like heat, low pH and others27. Exposure of the virions to susceptible cells was also reported Domain A Fusion loop to result in transitional epitope accessibility, which is suggestive of an q3 Domain B A early conformational change related to cell binding28. The residues in R68 B the E1 transitional epitopes map to domain III, in a region that is q2 F 2 2 58 E buried in the particle at the spike interfaces (Fig. 3c). The E2 transi- C D60 C′ tional epitope is in domain B, in a region contacting the E1 fusion loop 3 q2 T197 (Figs 2a and 3c). The fact that this surface becomes exposed indicates q2 that domain B moves out, thereby opening the groove and releasing q3 β-Ribbon the fusion loop. This observation is in agreement with the low pH Arch 2 2 structure reported in the accompanying paper3, in which domain B Central arch SINV N196 glycan is out of place and disordered in the crystal. Furthermore, the locations of the E1 transitional epitopes indicate formation of alternative inter- Figure 3 | Combination with cryo-EM data. a, The alphavirus T 5 4 spike contacts, indicating a possible allosteric transmission of the icosahedral surface glycoprotein shell. Atomic model of the 240 chikungunya ‘uncapping’ signal across the particle surface (see Supplementary E2–E1 heterodimers arranged as 80 spikes. E2 is coloured as in Fig. 1 and E1 is Movie). A detectable expansion of the alphavirus icosahedral particle sandy brown for clarity. b, Top view of the spikes. The 9 Å resolution cryo-EM map of the Sindbis virus (SINV) virion30 is represented as a grey transparent on exposure to low pH has indeed been reported29. surface. The T 5 4 icosahedral symmetry and quasi-symmetry axes are The low pH structure of the alphavirus E2–E1 heterodimer3 shows indicated. The black circle marks the close-up view of d. c, Side view of a spike. that domain B and half of the b-ribbon connector become disordered, Red spheres mark the transitional epitopes on E2 domain B and on E1 domain with domains A and C remaining in place. The missing half of the III, in a region buried at interspike contacts. Regions of the map not fitted in this b-ribbon, termed the acid sensitive region (ASR), is a region sandwiched work (stem region, transmembrane (TM) domains, capsid and viral between domain A of E2, E3 and E1 in the CHIKV structures (Sup- membrane) are indicated. d, Close-up of the spike top. The fusion loop plementary Fig. 5a). The E2–E1 interactions in this region are centred (labelled) is capped by E2 domain B. The extra density extending away of around a hydrogen bond between the side chains of E2 His 170 and E1 domain B corresponds to an N-linked glycan in SINV (attached to N196, corresponding to Thr 197 in CHIKV, labelled). In domain A, the side chains of Ser 57 (Fig. 4 and Supplementary Fig. 2), which are conserved in all Asp 60 and the buried Arg 68, shown in yellow sticks and labelled, make a mosquito-borne alphaviruses (Supplementary Fig. 1b, c). In the imma- stabilizing salt bridge. Position 58, a virulence determinant (see Supplementary ture particle, domain A and E3, which are tethered by the N linker (see Tables 6 and 7), is marked in red. Supplementary Information), clamp the b-ribbon in place until furin cleaves the tether (Fig. 4b). The CHIKV structures thus explain the resistance of immature alphavirus particles to activation by low pH. Furin loop Furin loop a b E3 N linker * N linker Domain B Domain A Arch2 90° * E3 Central arch *c e * *e Arch2 * Arch2 H170 S57 d c H170 b S57 d FL FL Domain II Domain II 50 Å Figure 4 | p62 maturation and low pH conformational transition. triangles under the scale bar indicate the slab used for b. b, Clamping by E3 and a, b, Interactions of the ASR (see Supplementary Information) of E2 with E1 domain A on the ASR. View at 90 degrees from a, as indicated. The N linker is domain II, centred on the hydrogen bond (thick black line) between E2 His 170 highlighted with yellow dots. Cleavage by furin (red arrow) releases the and E1 Ser 57 (labelled). Black and white asterisks indicate the ends of visible constraint, allowing the dislodgement of the ASR on low pH destabilization of density in the low pH structure of SINV E2–E13. a, Side view. The tip of E1 its contacts with E1 (movement indicated by the purple transparent arrow). domain II and the fusion loop projecting away from the paper with Trp 89 is in Red asterisks delimitate the region that becomes disordered on furin cleavage. white. Domain A is in the background, behind the N linker. The two solid black 2 D E C E M B E R 2 0 1 0 | VO L 4 6 8 | N AT U R E | 7 1 1 ©2010 Macmillan Publishers Limited. All rights reserved  RESEARCH LETTER This analysis is corroborated by experimental data mapping virus rescue 15. Marsh, M. & Helenius, A. Virus entry into animal cells. Adv. Virus Res. 36, 107–151 (1989). mutations in p62 to the ASR (Supplementary Table 5). 16. Wahlberg, J. M., Boere, W. A. & Garoff, H. The heterodimeric association between In conclusion, this study reveals the organization of the mature and the membrane proteins of Semliki Forest virus changes its sensitivity to low pH immature alphavirus surface glycoprotein complexes, the nature of the during virus maturation. J. Virol. 63, 4991–4997 (1989). 17. Kielian, M. & Helenius, A. pH-induced alterations in the fusogenic spike protein of initial activating transition when exposed to low pH and the inbuilt Semliki Forest virus. J. Cell Biol. 101, 2284–2291 (1985). mechanism to prevent this change from taking place prematurely. 18. Wahlberg, J. M., Bron, R., Wilschut, J. & Garoff, H. Membrane fusion of Semliki Together with the accompanying paper3, the structures show that the Forest virus involves homotrimers of the fusion protein. J. Virol. 66, 7309–7318 (1992). first step of the alphavirus fusogenic transition is removal of the domain 19. Lescar, J. et al. The fusion glycoprotein shell of Semliki Forest virus: an icosahedral B cap covering the fusion loop, without full dissociation of the E2–E1 assembly primed for fusogenic activation at endosomal pH. Cell 105, 137–148 heterodimer. The structures also reveal the organization of the indi- (2001). 20. Gibbons, D. L. et al. Conformational change and protein–protein interactions of the vidual immunoglobulin-like domains of E2 that are responsible for fusion protein of Semliki Forest virus. Nature 427, 320–325 (2004). receptor interactions, carrying important determinants of virulence 21. Kielian, M. & Rey, F. A. Virus membrane-fusion proteins: more than one way to and mosquito vector range. These domains also carry the epitopes make a hairpin. Nature Rev. Microbiol. 4, 67–76 (2006). targeted by neutralizing antibodies, opening the way for vaccine design 22. Rey, F. A. et al. The envelope glycoprotein from tick-borne encephalitis virus at 2 Å resolution. Nature 375, 291–298 (1995). by immunization against alphaviruses with small recombinant proteins 23. Lobigs, M., Zhao, H. X. & Garoff, H. Function of Semliki Forest virus E3 peptide in instead of whole particles. virus assembly: replacement of E3 with an artificial signal peptide abolishes spike heterodimerization and surface expression of E1. J. Virol. 64, 4346–4355 (1990). 24. Tsetsarkin, K. A. et al. Epistatic roles of E2 glycoprotein mutations in adaption of METHODS SUMMARY chikungunya virus to Aedes albopictus and Ae. aegypti mosquitoes. PLoS ONE 4, Purification, crystallization and X-ray diffraction data collection. The ectodo- e6835 (2009). mains of CHIKV-115 p62 and E1 joined with a glycine serine linker were 25. Pierro, D. J., Powers, E. L. & Olson, K. E. Genetic determinants of Sindbis virus expressed and secreted from stably transfected S2 cells as a strep-tagged (IBA, mosquito infection are associated with a highly conserved alphavirus and http://www.iba-go.com/) recombinant protein. The protein was purified by affinity flavivirus envelope sequence. J. Virol. 82, 2966–2974 (2008). 26. Meyer, W. J. & Johnston, R. E. Structural rearrangement of infecting Sindbis virions chromatography and limited proteolysis was used to cleave the linker. Diffraction at the cell surface: mapping of newly accessible epitopes. J. Virol. 67, 5117–5125 data from native and several heavy-atom-derivative crystals were collected at syn- (1993). chrotron sources (Supplementary Table 1a, b). 27. Meyer, W. J. et al. Conformational alteration of Sindbis virion glycoproteins induced Structure determination and refinement. The structure was determined using a by heat, reducing agents, or low pH. J. Virol. 66, 3504–3513 (1992). combination of molecular replacement with the available structure of alphavirus 28. Flynn, D. C., Meyer, W. J., Mackenzie, J. M. Jr & Johnston, R. E. A conformational E1, experimental phasing with heavy atoms, density modification, automatic change in Sindbis virus glycoproteins E1 and E2 is detected at the plasma membrane as a consequence of early virus-cell interaction. J. Virol. 64, 3643–3653 model building and multi-crystal averaging. Atomic models were refined against (1990). diffraction data from five crystal forms (Supplementary Table 1c). 29. Wu, S. R. et al. The dynamic envelope of a fusion class II virus. Prefusion stages of EM fitting. E2–E1 heterodimers were fitted in EM maps as described in Sup- Semliki Forest virus revealed by electron cryomicroscopy. J. Biol. Chem. 282, plementary Information. The heterodimer was initially fit as a single rigid body 6752–6762 (2007). and then was cut into five parts as defined in Supplementary Table 3. 30. Mukhopadhyay, S. et al. Mapping the structure and function of the E1 and E2 glycoproteins in alphaviruses. Structure 14, 63–73 (2006). Full Methods and any associated references are available in the online version of Supplementary Information is linked to the online version of the paper at the paper at www.nature.com/nature. www.nature.com/nature. Received 24 May; accepted 5 October 2010. Acknowledgements We thank the CHIKV task force at Institut Pasteur, in particular the group of F. Tangy and the staff of platform PF8 for the CHIKV complementary DNA; 1. Her, Z., Kam, Y. W., Lin, R. T. & Ng, L. F. Chikungunya: a bending reality. Microbes A. Haouz of PF6 for crystallogenesis; the staff of synchrotron beamlines PROXIMA 1 at Infect. 11, 1165–1176 (2009). Soleil, ID23-eh2 at the European Synchrotron Radiation Facility and PX-I at the Swiss 2. Schwartz, O. & Albert, M. L. Biology and pathogenesis of chikungunya virus. Nature Light Source; M. Rossmann and Y. Sun for providing the 16 Å cryo-EM map of CHIKV Rev. Microbiol. 8, 491–500 (2010). virion-like particles and for sharing the coordinates and manuscript of the low pH 3. Li, L., Jose, J., Xiang, Y., Kuhn, R. J. & Rossmann, M. G. Structural changes of structure of the SINV E1–E2 heterodimer before publication; and members of the F.A.R. envelope proteins during alphavirus fusion. Nature doi:10.1038/nature09546 laboratory for help during data collection. J.E.V. was supported by a Marie Curie (this issue). fellowship through the European Union Research Traning Network program 4. Halstead, S. B. in Pediatric Infectious Diseases (eds Feigin, R. & Cherry, J.) ‘‘Intrapath’’. This work was funded in part by the French ‘Agence Nationale de la 2178–2183 (Saunders, 2004). Recherche’ grant DENtry in the program ‘Microbiologie, Infections et Immunité’, by 5. Robinson, M. C. An epidemic of virus disease in southern province, Tanganyika Merck-Serono, by the Pediatric Dengue Vaccine Initiative and by the Institut Pasteur territory, in 1952–1953 I. Clinical features. Trans. R. Soc. Trop. Med. Hyg. 49, 28–32 program PTR201 CHIKV to F.A.R.. (1955). Author Contributions J.E.V. made the constructs, produced and purified the protein, 6. Johnson, F. Notes on Kimakonde. Bull. Sch. Orient. Studies 2, 417–466 (1922). grew the crystals and participated in diffraction data collection; analysed the literature 7. Carey, D. E. Chikungunya and dengue: a case of mistaken identity? J. Hist. Med. and prepared tables and figures. M.-C.V. carried out most of the various Allied Sci. 26, 243–262 (1971). crystallographic refinements and prepared the figures, S.D. carried out the fitting into 8. Schuffenecker, I. et al. Genome microevolution of chikungunya viruses causing the the cryo-EM maps of various alphavirus particles and prepared the figures, C.G.-B. and Indian Ocean outbreak. PLoS Med. 3, e263 (2006). E.C. participated in optimizing protein production in large scale for crystal trials, C.V. 9. Tsetsarkin, K. A., Vanlandingham, D. L., McGee, C. E. & Higgs, S. A single mutation in and G.B. participated in data processing and in the structure determination. A.T. carried chikungunya virus affects vector specificity and epidemic potential. PLoS Pathog. out specific data collection strategies to improve the signal to noise to extract 3, e201 (2007). anomalous signal for phasing; F.A.R. conceived the experiments and wrote the 10. Vazeille, M. et al. Two chikungunya isolates from the outbreak of La Reunion manuscript. (Indian Ocean) exhibit different patterns of infection in the mosquito, aedes albopictus. PLoS ONE 2, e1168 (2007). Author Information Structure factors and coordinates for the structures of the p62–E1 11. Enserink, M. Entomology: a mosquito goes global. Science 320, 864–866 (2008). (crystal IO), E3–E2–E1sp (crystal MM), E3–E2–E1f (crystal MO1), E3–E2–E1t (crystal 12. Akahata, W. et al. A virus-like particle vaccine for epidemic chikungunya virus MO2) and Os2-E3–E2–E1t (crystal MO3) complexes have been deposited in the Protein protects nonhuman primates against infection. Nature Med. 16, 334–338 (2010). Data Bank under accession codes 3N40, 3N41, 3N42, 3N43 and 3N44, respectively. 13. Salminen, A. et al. Membrane fusion process of Semliki Forest virus. II. Cleavage- The coordinates of the molecules fitted in the cryo-EM 3D reconstructions of SINV and dependent reorganization of the spike protein complex controls virus entry. J. Cell SFV particles were deposited under accession codes 2XFB and 2XFC, respectively. Biol. 116, 349–357 (1992). Reprints and permissions information is available at www.nature.com/reprints. The 14. Ziemiecki, A., Garoff, H. & Simons, K. Formation of the Semliki Forest virus authors declare no competing financial interests. Readers are welcome to comment on membrane glycoprotein complexes in the infected cell. J. Gen. Virol. 50, 111–123 the online version of this article at www.nature.com/nature. Correspondence and (1980). requests for materials should be addressed to F.A.R. ([email protected]). 7 1 2 | N AT U R E | VO L 4 6 8 | 2 D E C E M B E R 2 0 1 0 ©2010 Macmillan Publishers Limited. All rights reserved  LETTER RESEARCH METHODS these crystals showed that there had been cleavage by an unknown contaminant Recombinant protein production and purification. We produced the recom- protease in the region of the linker during the crystallization process (crystal E3– binant p62–E1 and E3–E2–E1 complexes from the clinical isolate 05-115 (ref. 31) E2–E1sp, where the ‘sp’ subscript stands for ‘spontaneous cleavage’, meaning un- in Drosophila melanogaster Schneider 2 (S2) cells with the construct outlined in intended cleavage). The controlled chymotrypsin treatment outlined earlier was Fig. 1a, bottom-right panel. A 19-residue linker, with sequence (GGGGS)4 (omit- then introduced into the purification protocol, resulting in the growth of ortho- ting the last Ser residue) connected the C terminus of the p62 ectodomain to the N rhombic crystals. These include both the immature form obtained from the furin terminus of E1, thus bypassing the p62 transmembrane region and the 6K protein site mutant (crystal p62–E1, coded IO for ‘immature orthorhombic’; see Sup- (Supplementary Fig. 1a). The actual construct contained amino acids 1–421 of p62 plementary Table 1) as well as the mature forms; cleaved naturally by furin during (the last one corresponding to E2 residue 361) and 1–412 of E1 (that is, all of the expression for the wild type (crystal E3–E2–E1f, code MO1, for ‘mature ortho- ectodomain, stopping right at the beginning of the transmembrane segment for rhombic form 1’) or by trypsin treatment of the p62–E1 complex as described both proteins; see Supplementary Fig. 1a, b). We tried a shorter construct in earlier (crystal E3–E2–E1t, code MO2). The latter crystals were used to screen parallel, spanning p62 residues 1–404 (the end corresponding to E2 position for heavy atom derivatives to obtain experimental phases. Useful phase informa- 340) and E1 1–392 joined with the same linker, but we did not get diffraction- tion was recovered from osmate derivatives of the E3–E2–E1t crystals obtained by quality crystals from the resulting protein. For expression in Drosophila cells, we soaking in 1 mM or in 10 mM potassium osmate (K2OsO4) solutions overnight. used a modified version of the pMRBiP/V5 HisA plasmid (Invitrogen) in which a Also, a rare-earth derivative was obtained by soaking in a 50 mM holmium chloride double strep-tag (IBA, http://www.iba-go.com/) replaces the C-terminal His tag32. (HoCl3) solution for 2 h. All of the crystals were flash frozen in liquid nitrogen using In addition, the Drosophila BiP signal sequence was replaced by the authentic p62 25% ethylene glycol as cryoprotectant. Diffraction data were collected at signal sequence, which retracts from the membrane to make part of the E3 domain the European synchrotrons listed in Supplementary Table 1, and were processed of p62 during folding (Supplementary Fig. 1a). The secreted protein was the with autoPROC (version 0.5.3; C. Vonrhein et al., unpublished) using XDS33, mature E3–E2–E1 complex, resulting from maturation of p62 by Drosophila furin. POINTLESS34, SCALA and TRUNCATE from the Collaborative Computational To produce the immature complex we introduced a mutation at the furin site, Project, No. 4 (CCP4) suite35. Data collection statistics for native and heavy atom sequence 61-RQRR#S-65 (the down arrow indicating the site of cleavage) mutated derivative crystals are presented in the Supplementary Table 1a. into 61-RQRES-65 to avoid furin cleavage. In the end, we found it more conveni- Structure determination. Initial phases were obtained by molecular replacement ent to use the latter construct, and cleave the immature complex with trypsin at the (MR) using AMoRe36 with the structure of monomeric glycoprotein E1 from SFV remaining basic residues of the mutated furin site to obtain the mature E3–E2–E1 (PDB code 2ALA)37. The model used successfully for MR was a trimmed version, complex when needed (both structures, furin-cleaved and trypsin-cleaved, were lacking the regions 52–109 and 217–238 (that is, the tip of domain II). The MR determined in this work, and they are identical). solution did not produce interpretable density for the remaining p62 part. These Stable transfectant S2 cell lines were generated with these recombinant plas- initial phases, however, were accurate enough to locate several heavy atom sites in mids. One-litre cultures in logarithmic phase growth (Insect Express Media) were anomalous difference Fourier maps for the 10 mM osmate derivative to 3.7 Å routinely induced with 500 mM CuSO4 to secrete protein for 1 week. The super- (Os1-E3–E2–E1t) (Supplementary Table 1a). Those sites, together with the phases natants were then collected, concentrated and filtered with 0.2 mm cutoff mem- from the partial molecular replacement model, were refined and used for phasing brane and adjusted to pH 8 with Tris buffer containing avidin at 15 mg ml21. The in SHARP38. The log-likelihood gradient maps were used to complement and protein was purified by affinity chromatography using 4 ml StrepTactin columns correct the heavy atom model. (IBA). After loading the protein, the column was washed with 100 mM Tris pH Density modification with SOLOMON39 produced electron density maps that 8.0, 150 mM NaCl, 1 mM EDTA, and the sample was eluted using 2.5 mM desthio- allowed the interpretation of the missing part of E1. The density in the p62 region biotin. We adjusted the overall purification protocol after the initial crystallization. could be interpreted in part by BUCCANEER40 as polyalanine fragments. This Screening results showed that cleavage of the artificial linker introduced in gave a more complete model covering not only most of E1 but also significant between p62 and E1 was necessary to obtain diffraction quality crystals. We found portions of p62. The improved phases obtained by combining the phase informa- that the best crystals were obtained with protein digested with chymotrypsin, tion from the intermediate partial models with the experimental phases were used which cleaved in the region rich in aromatic residues at the end of p62 (around to determine the heavy atom sites in the other heavy atom derivative data sets position 425, which corresponds to 360 in E2 numbering; Supplementary Fig. 1b), (Os2-E3–E2–E1t and Ho-E3–E2–E1t). Because the various data sets were not right before the linker. In our final protocol, the sample eluted from the isomorphous, showing significant variation in cell dimensions (up to 8 Å in the StrepTactin column was pooled and treated with chymotrypsin (Roche) at a mass c axis for Os1-E3–E2–E1t) (Supplementary Table 1a), the phase information from ratio of 1:5,000 enzyme:protein overnight at room temperature (20 uC). After different data sets was combined by using multi-crystal averaging in DMMULTI41. stopping the reaction by the addition of phenylmethanesulphonyl fluoride Averaging masks were based on the best partial model available at each step. Multi- (PMSF), the protein was desalted into 10 mM Tris pH 8.0, 10 mM NaCl crystal averaging was done using the monoclinic crystal form (map calculated with (HiTrap Desalting Column 5 ml, GE Healthcare) to perform an ion exchange MR phases) with the orthorhombic crystal forms (using SIRAS data and phases purification step. For this purpose, the sample was loaded onto a MonoQ 5/50 based on the Ho-E3–E2–E1t and p62–E1 data sets, anisotropically corrected column (GE Healthcare). A salt gradient from 10 to 200 mM NaCl was applied, within SHARP). The averaged map resulting from this multi-averaging step and the protein eluted as several overlapping peaks, which were collected in sepa- revealed the missing parts of the structure, which were built manually with the rate fractions, concentrated on a vivaspin 30K cutoff, and applied to a gel filtration program COOT42. After rebuilding the tip of E1 domain II, including the fusion Superdex 10/300 column equilibrated in 20 mM Tris pH 8.0, 100 mM NaCl. The loop, the N-terminal part of E3 and the furin loop, the structure obtained was gel filtration profile showed that the protein from the various peaks of the ion refined with BUSTER43 against the p62–E1 data at 2.17 Å. The refined p62–E1 exchange chromatogram eluted at a similar volume, corresponding to a molecular structure was then used in PHASER to obtain a molecular replacement solution weight of about 90 kD. The protein from this single peak was concentrated to about for the other crystals. All the structures were refined by applying nine ‘TLS’ groups, 6 mg ml21 for crystallization. Similar crystals were obtained from samples belong- describing the translation, libration and screw-rotation displacements (TLS44) per ing to different peaks in the ion exchange chromatogram. In cases where the complex. In the end, five crystal structures were refined: p62–E1 (IO), E3–E2–E1f mature form was obtained by trypsin treatment of the mutated p62–E1 hetero- (MO1), E3–E2–E1t (MO2), E3–E2–E1sp (MM) and Os2-E3–E2–E1t (MO3). The dimer, cleavage of the linker was done first, and after addition of PMSF to stop the position and interactions in the structure of the heavy atoms used for phasing is chymotrypsin reaction, desalting was done to remove excess PMSF and a second listed in Supplementary Table 1b, and refinement statistics for the five structures step of proteolysis was carried out with trypsin (Roche) also at a mass ratio of are presented in Supplementary Table 1c. The superposition of the five refined 1:5,000 enzyme:protein overnight at room temperature before the final purifica- models shows that there is variability at interdomain contacts (Supplementary Fig. 6a). tion steps by ion exchange and size exclusion chromatography described earlier. Fitting the crystallographic model into the cryo-EM density. The atomic model Crystallization and data collection. The crystals were grown by vapour diffusion of the E2–E1 heterodimers was extracted from the p62–E1 and E3–E2–E1sp crystal in hanging drops at 20 uC my mixing 1 ml of protein at a concentration of 5–7 mg structures (codes IO and MM, respectively; Supplementary Table 1) and were ml21 in 20 mM Tris pH 8.0, 100 mM NaCl with 1 ml of well solution containing fitted into the cryo-EM maps of SINV (Electron Microscopy Data Bank 8–12% PEG4K, 100 mM NaAcetate, 100 mM HEPES pH 7.1–7.5. The original (EMDB) code EMD-1121; ref. 45), SFV (EMDB code EMD-1015; ref. 46) and crystals were clusters of plates, and several successive seeding steps were used to CHIKV47 by maximizing the correlation between structure factors calculated from obtain single crystals large enough for diffraction data collection (thin plates, the cryo-EM maps and from the atomic model with the program URO48. The 200 3 100 3 5 mm3). Monoclinic crystals were obtained initially, coded as MM form CHIKV three-dimensional reconstruction used was a 16 Å cryo-EM map pro- for ‘mature monoclinic’ form (see Supplementary Table 1). Reproducing these vided by S. Sun and M. Rossmann. The procedure takes into account amplitude crystals was difficult, until SDS–polyacrylamide gel electrophoresis analysis together and phase information from the map and the T 5 4 symmetry of the cryo-EM with mass spectrometry and N-terminal sequencing of the protein recovered from reconstruction. The E2–E1 heterodimers refined in the IO and MM crystals, which ©2010 Macmillan Publishers Limited. All rights reserved  RESEARCH LETTER adopt the most extreme conformations among the five refined structures, were in Supplementary Fig. 1c, bottom-right panel) in each of the particles of the three fitted. The fits were done first as a single rigid body and were then improved using alphaviruses used (SINV, SFV and CHIKV) is shown in Supplementary Fig. 6c. several rigid bodies (five for SFV and SINV, and three for CHIKV) as defined in This figure shows that the adjustments between domains are similar to the varia- Supplementary Table 3. The SFV cryo-EM map was rescaled by a factor of 1.04 tions observed in different crystal forms, highlighting a natural flexibility of the after comparison with both CHIKV and SINV maps, to bring the viral membranes E2–E1 heterodimer, which is used to adjust to the quasi-equivalent position in the to a common radius. This step was done because fitting in the map without virus particle. rescaling led to small clashes at the interspike contacts, which were released in the expanded version. A sphere of a radius of 260 Å, 250 Å and 255 Å respectively 31. Schuffenecker, I. et al. Genome microevolution of chikungunya viruses causing the Indian Ocean outbreak. PLoS Med. 3, e263 (2006). for SINV, SFV (before rescaling) and CHIKV was used to eliminate density 32. Krey, T. et al. The disulfide bonds in glycoprotein E2 of hepatitis C virus reveal the belonging to the transmembrane regions and the nucleocapsid from the calcula- tertiary organization of the molecule. PLoS Pathog. 6, e1000762 (2010). tion, leaving only density corresponding to the alphavirus glycoprotein shell of the 33. Kabsch, W. Integration, scaling, space-group assignment and post-refinement. three different virions. The URO statistics resulting from the fitting are listed in Acta Crystallogr. D 66, 133–144 (2010). Supplementary Table 3a, whereas Supplementary Table 3b provides the surface 34. Evans, P. R. Scaling and assessment of data quality. Acta Crystallogr. D 62, 72–82 area buried in the various contacts in the particle between the fitted atomic models, (2005). together with the contacts observed within the E3–E2–E1 heterotrimer in the 35. Collaborative Computational Project Number 4.. The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D 50, 760–763 (1994). crystallographic model. The contacts on the spike are the interfaces between 36. Navaza, J. Implementation of molecular replacement in AMoRe. Acta Crystallogr. D heterodimers within a spike and across spikes on the virion (as defined in the 57, 1367–1372 (2001). key in the bottom-right panel of Supplementary Fig. 1c). Supplementary Table 3c 37. Roussel, A. et al. Structure and interactions at the viral surface of the envelope shows the root mean square deviation (r.m.s.d) after superposing the atomic protein E1 of Semliki Forest virus. Structure 14, 75–86 (2006). models fitted independently in the four positions of the T 5 4 icosahedral asym- 38. Bricogne, G. et al. Generation, representation and flow of phase information in metric unit of the particles, and on the different viruses. For comparison, the structure determination: recent developments in and around SHARP 2.0. Acta Crystallogr. D 59, 2023–2030 (2003). r.m.s.d. after superposition of the model refined in the different crystals is also 39. Abrahams, J. P. & Leslie, A. G. Methods used in the structure determination of provided in the same table. Supplementary Table 3d shows the Ca–Ca distance bovine mitochondrial F1 ATPase. Acta Crystallogr. D 52, 30–42 (1996). (normally 3.8 Å) resulting between the C termini and N termini of subsequent 40. Cowtan, K. The Buccaneer software for automated model building. 1. Tracing segments of the polypeptide chain that were refined in different rigid bodies, to protein chains. Acta Crystallogr. D 62, 1002–1011 (2006). give an idea of the distortion caused to the model (most of the breaks can be 41. Cowtan, K. ‘dm’: an automated procedure for phase improvement by density reconnected manually without major rearrangement of the polypeptide chain). modification. Joint CCP4 ESF-EACBM Newsl. Protein Crystallogr. 31, 34–38, (1994). 42. Emsley, P. et al. Features and development of Coot. Acta Crystallogr. D 66, 486–501 Finally, Supplementary Fig. 6 shows a comparison of the inter-domain variations, (2010). reflecting inherent flexibility of the E2–E1 heterodimer. In all three panels, the Ca 43. Bricogne, G. et al. BUSTER version 2.9. (Global Phasing, 2010). atoms of the first portion of domain II (closest to domain I) of E1 were superposed 44. Painter, J. & Merritt, E. A. A molecular viewer for the analysis of TLS rigid-body and displayed so that the variation in orientation of the adjacent domains of E1 motion in macromolecules. Acta Crystallogr. D 61, 465–471 (2005). and E2 can be evaluated visually. Supplementary Fig. 6a shows this superposition 45. Mukhopadhyay, S. et al. Mapping the structure and function of the E1 and E2 after refining the atomic E2–E1 atomic models against the various crystals listed in glycoproteins in alphaviruses. Structure 14, 63–73 (2006). Table 1c. Supplementary Fig. 6b shows this same superposition with the model 46. Mancini, E. J. et al. Cryo-electron microscopy reveals the functional organization of an enveloped virus, Semliki Forest virus. Mol. Cell 5, 255–266 (2000). resulting from fitting the E2–E1 heterodimer as five rigid bodies into the four 47. Akahata, W. et al. A virus-like particle vaccine for epidemic Chikungunya virus icosahedrally independent locations of the T 5 4 surface lattice of the SINV cryo- protects nonhuman primates against infection. Nature Med. 16, 334–338 (2010). EM map and superposed in the same way. Superposition of the atomic models 48. Navaza, J. et al. On the fitting of model electron densities into EM reconstructions: a resulting from fitting at positions P of the T 5 4 icosahedral lattice (see definition reciprocal-space formulation. Acta Crystallogr. D 58, 1820–1825 (2002). ©2010 Macmillan Publishers Limited. All rights reserved