Alphaviruses generate invaginations in the plasma membrane or cytoplasmic vacuoles of host cells, creating small (around 50 nm in diameter) lipid baskets known as spherules2. Spherules have a small aperture that connects with the external cytosol, which is the only way out for the newly synthetized viral RNA and the only way in for the metabolites and proteins that are required for RNA synthesis. The extruded viral RNAs are capped and polyadenylated, ready for translation of the viral proteins by the host. This is achieved by viral replication complexes that assemble onto cellular membranes, promoting spherule formation and synthesizing viral RNAs in a coordinated fashion.

Alphaviruses have a replication complex that contains four non-structural proteins (nsP1, nsP2, nsP3 and nsP4), which are produced following cleavage of the viral polyprotein precursor by a viral protease in nsP2. nsP1, nsP2, nsP3 and nsP4 are essential for replication9,10. nsP1 is the only one of these proteins that is known to interact directly with the membrane, and it displays two activities that are necessary for the capping of viral RNA: S-adenosyl-l-methionine (SAM)-dependent methyltransferase (MTase) and m7GTP transferase (GTase) activities11,12,13,14, both of which are dependent on membrane binding15. Here we show the structure of the nsP1 complex, which provides a structural basis for these phenomena and challenges biochemical data that have accumulated over decades on positive-sense single-stranded RNA viruses replicating in membrane-associated viral factories.

Assembly of nsP1 into active dodecameric pores

We expressed nsP1 recombinantly in Escherichia coli and in insect cells16, and recovered inactive monomeric and active membrane-associated nsP1, respectively. Detergent screening enabled us to solubilize and purify active ring-shaped complexes of nsP1 from insect cells and inactive monomeric protein, which indicates that nsP1 of chikungunya virus (CHIKV) needs to be assembled as oligomers for activity (Extended Data Fig. 1, Methods). We determined the structure of the complex using single-particle cryo-electron microscopy (cryo-EM). The sorting of particles into two-dimensional (2D) and three-dimensional (3D) class averages was consistent with two volumes, which correspond to dodecameric single and double rings of nsP1: the structures of these rings were determined at a resolution of 2.6 Å and 2.9 Å, respectively (Extended Data Figs. 2, 3, Extended Data Table 1). The absence of contacts between the interface of the double rings confirms that the single ring is the biological assembly (Extended Data Fig. 4). Twelve molecules of nsP1 assemble in the ring with C12 symmetry (Fig. 1a). The rings are 18.6 nm in diameter and 7 nm in height, and have a central aperture that is 7 nm wide (Fig. 1b, c). Three regions are well-defined in the complex: the crown, the waist and the membrane-binding skirt, which together confer a bodice-like shape to the complex (Fig. 1c). The crown is formed from the capping domains of the 12 subunits (as discussed in ‘Atomic structure of CHIKV nsP1 protein’), generating a cone-shaped inner chamber that is negatively charged, about 3.7 nm in depth and has a diameter of 14 nm at the top that narrows to 7 nm at the bottom. The top of the crown is lined with positively charged pockets that correspond to the active sites of the capping domain. Under the crown, the 7-nm-wide waist region defines a pore that is 2 nm in depth as well as a positively charged region outside the complex. The inner walls of the pore have a neutral charge and would allow for the passage of globular proteins smaller than about 70–90 kDa in size. Under the waist, spikes from the membrane-binding skirt project into the detergent micelles at an angle of around 35° from the equatorial plane of the pore, generating a concave and positively charged surface that extends from the bottom of the complex to the lower external waist (with the exception of the hydrophobic tips of the spikes). Tomographic reconstructions of spherule necks, performed in a nodavirus (Flock House virus (FHV)) at low resolution, maintain dimensions and architecture similar to those shown by the nsP1 complex9 (Fig. 1d, Extended Data Fig. 5). In conclusion, the nsP1 complex forms enzymatically active rings that are monotopically associated with the membrane within the spherule necks, and which gate access to the organelle. This is the foundation of the generation of spherule-like viral replication factories.

Fig. 1: Cryo-EM structure of the spherule–pore complex of CHIKV nsP1.
figure 1

a, Ribbon representation of the nsP1 complex from its cytosolic (top) view. The 12-fold symmetry is indicated by numbering; each protomer is shown in a different colour. b, Surface representation of the top cytoplasmic view (top) and membrane-bound side (bottom). The surface electrostatics were calculated using APBS ( from −5 kT/e to 5 kT/e, and are represented by the colour code of the bottom bar. The diameters of the ring and the pore aperture are indicated. c, Same representation as in b, of the lateral view of the complex (left) and a section through the broader part (right). The three regions (crown, waist and skirt) are delimited by black lines. The dimensions of the height, the cone upper aperture, the pore height and the bottom skirt diameter are indicated. The membrane-binding spikes are visible at the bottom of the skirt. The detergent-binding interface (substituting the membrane) is indicated by a thick dashed line. The projection angle of the spikes with respect to the pore is also indicated. d, Lateral view of tomographic reconstructions of FHV spherule necks4 in which the CHIKV nsP1 complex is superposed (Extended Data Fig. 5). nsP1 is shown coloured by protomer inside the FHV volume (represented as a transparent grey surface). e, Schematic of the structural regions of nsP1, indicating the residues that delimit each region (CHIKV numbering). The green fragment (NT) is the N-terminal extension of the capping domain. The grey fragment corresponds to the C-terminal residues that are disordered in the structure (Extended Data Fig. 6). f, Two-side view of the nsP1 protomer structure, represented as a cartoon following the same colour code as in e. All secondary structures and the Zn atom are labelled, as are the corresponding regions of the crown, waist and skirt of the ring. The missing loop between αi and αh is indicated by blue dashed lines. The circled C and N represent the C and N termini, respectively.

Atomic structure of CHIKV nsP1 protein

We built the atomic model of nsP1 de novo: it includes residues 3–471 out of 535, with density gaps at 364–376 and 450–458. No substantial differences were found between protomers of the double and single rings. nsP1 folds as two interconnected domains: the capping domain (located in the crown of the complex) and the ring-aperture membrane-binding and oligomerization (RAMBO) domain, which defines the waist and skirt regions of the complex (Fig. 1e, f). The capping domain (residues 1–294 and 459–472) exhibits high structural similarity with other SAM-dependent MTases: the DALI protein-structure comparison server17—using the complete Protein Data Bank (PDB) database—identifies MTases of Encephalitozoon cuniculi (PDB code 1RI1), vaccinia virus (PDB code 6RFL) and human (PDB code 5E8J) (among others) as being similar to CHIKV nsP1, with respective Z-scores of 10.1, 10.0 and 9.6, and root mean square deviations of 3.2 Å for all of them, including—respectively—188, 190 and 192 residues in the structural alignments (Extended Data Fig. 6). All of these MTases exhibit variations of a canonical SAM-dependent fold that comprises a core β-sheet of seven β-strands (β1 to β7) flanked by α-helices. In nsP1, the core β-sheet is modified through substitution of β3 by the α-helix αC (which forms an α-bundle with αB and αk; upper- and lower-case letters denote conserved secondary structures among MTases and specific secondary structures of CHIKV MTase, respectively), the incorporation of an additional β-strand (β6′) at the outer edge of the sheet and the inclusion of a Zn-binding site formed by the tetrahedral arrangement of the side chains of residues H79, E129, C134 and C141 (Extended Data Fig. 3a). The nsP1 N terminus is extended by a β–α–β motif (β1′–αa–β2′). An additional, smaller β-sheet insertion (β8 to β11) is conserved in vaccinia virus, human and E. cuniculi MTases18,19, but is absent in other positive-stranded RNA viruses such as flavivirus MTases20. This insertion is involved in the binding of a GTP acceptor in the methyltransferase reaction. Within this region, nsP1 has a 35-amino-acid-long loop insertion between β9 and β10 (residues 201–236) that is critical for membrane binding and oligomerization (which we term membrane binding and oligomerization loop 1 (MBO loop 1)) (Extended Data Fig. 6a). We could assign weak density, consistent with an α-helix (αk), to residues 462–472 at the C terminus of nsP1 (Extended Data Fig. 3b). Beyond this point, density could not be resolved, which indicates that the missing C-terminal 60 residues are disordered and projected towards the top of the crown.

The RAMBO domain (residues 295–450) folds in two β-sheets of antiparallel strands (β13–β15 and β12–β16), forming the platform on which the capping domain sits. The α-bundle that follows these strands comprises three α-helices (αh, αi and αj), which form the inner walls of the pore. A loop between αh and αi (residues 365–375) is not visible, which implies that it is highly flexible and deployed towards the cone area. Infections with temperature-sensitive mutants of the Sindbis virus RNA-dependent RNA polymerase nsP4—which prevent the synthesis of negative-strand RNA under non-permissive temperatures—could be compensated by mutations in nsP1 (T349K or T349L and N374H or N374I) (CHIKV T351 and N375) that are, respectively, located at the base and in the disordered loop at the top of the α-bundle21,22 (Extended Data Fig. 6d). This suggests that nsP4 may interact with both the inner and the outer sides of the pore. Beneath the complex, a long loop (which we term the MBO loop 2) and a pair of long antiparallel strands (β13 and β14 (which we term the MBO β-turn)) are involved in membrane binding (as discussed in ‘nsP1 oligomerization defines membrane binding’).

In conclusion, the nsP1 capping enzyme is derived from a classic SAM-dependent MTase fold with modifications that allow for the capping reaction, multimerization and membrane binding. The modifications include novel insertions and extensions at the N terminus and the C terminus of the MTase fold. Sequence similarity shows that this folding is conserved among all alphaviruses (Extended Data Figs. 6d, 7).

nsP1 oligomerization defines membrane binding

The structure shows that oligomerization is mediated by contacts that involve the capping and, mainly, the RAMBO domain. Our analysis of the complex interfaces using the PISA server23 reveals an extensive interface that engages 84 amino acid residues in contacts with the adjacent nsP1n − 1 and 92 residues in contacts with nsP1n + 1 subunits, burying 3,282 Å2 and 3,213 Å2 of protein surface, respectively. Each of the 12 interfaces includes 49 hydrogen bonds and 19 salt bridges, (Extended Data Figs. 7, 8, Supplementary Table 1). Altogether this network of interactions buries 29.3% of the nsP1 surface area (reaching 77,940 Å2 for the whole ring), and engages most of the nsP1 secondary structures. As a consequence, oligomerization allosterically activates the enzyme by stabilizing the conformation of the capping domain and generating a bilobular catalytic pocket in which the GTP-binding lobe faces the interior, and the SAM-binding lobe faces the exterior, of the crown (Fig. 2a).

Fig. 2: nsP1 oligomerization and membrane binding are coupled processes.
figure 2

a, Left, the top view of the ring is represented as a cartoon and coloured as in Fig. 1f. The active site cavities are depicted as green meshes. Access to the GTP- and SAM-binding pockets in the active site from each side of the ring is indicated. The active site (dashed box) is magnified on the right; the view is indicated by the eye icon. The entrance of the GTP and the SAM to their respective binding pockets in the active-site cavity is indicated by arrows. The pocket is covered by the neighbouring n − 1 subunit (in grey). SAH and GTP molecules are superimposed from the structure of Ecm1 (PDB code 2RI1). b, Cartoon representation of the spike (dashed box), with magnified detail on the right. The n protomer is coloured as in a; the MBO β-turn and MBO loop 1 are coloured in red; and MBO loop 2 from the n − 1 protomer is coloured in green. The MBO loop residues are shown as sticks. The curved dashed line indicates the boundaries of the micelles found in the double-ring structure (Extended Data Fig. 4). A black arrow indicates the 10-Å length of the nsP1 fragment inserted into the FC12 micelles. c, MBO loop 2 represented as in b, with the density of the cryo-EM structure in blue mesh (drawn in PyMOL at 2σ). The FC12 detergent micelle boundary is indicated by a dashed line as in b. Details of the cysteine triad density are shown in Extended Data Fig. 3i–k. d, Representation of the MBO loop 1, as in c.

The most notable part of the interface is generated in the vicinity of the membrane by the MBO loop 2 from the RAMBO domain, which wraps under MBO loop 1 of the capping domain and the β-turn of the RAMBO domain of the nsP1n − 1 (Fig. 2b). These elements fold together, forming amphipathic membrane-binding spikes that penetrate about 10 Å into the detergent FC12 micelle. The tip of the spike is hydrophobic and includes residues 225LSIM228 from MBO loop 1 and 416TCCCLWA422 from nsPn + 1 MBO loop 2. The bases of the spikes create a positively charged belt that is continuous around the skirt, and which constitutes a platform for the binding of the negatively charged phospholipid heads. A triad of cysteines in MBO loop 2 are known to be palmitoylated24,25 in mammalian cells, but—consistent with the very low levels of palmitoylation detected in our insect-cell-expressed nsP1 (Extended Data Fig. 1g)—we do not observe acylation in the structure (Extended Data Fig. 3i–k). However, the cysteine triad appears in the structure at the end of the tip inside the micelles, consistent with the putative insertion of the palmitoyl moieties into the cellular membrane. Thus, palmitoylation is not essential for pore formation, which suggests that other eukaryotic cell factors must be required. This is consistent with previous studies that have demonstrated that palmitoylation is a determinant of nsP1 localization within the membrane, rather than of activity or membrane binding24. However, in combination, the approximately 17-Å-long palmitoyl chains and the approximately 10-Å-long inserted tip can reach a depth of around 2.7 nm into the lipid bilayer (the thickness of which ranges from about 2.5 to 3.5 nm). This represents a very strong monotopic interaction with the lipid bilayer. The well-defined electron density maps in the bases of the spikes indicates that the spikes are rigidly projected from the complex (Fig. 2c, d). The strength of the membrane binding, the rigidity of the spikes and the curvature of the positively charged skirt edges—which attract the membrane phospholipid heads—can potentially induce the marked membrane bending that is observed in spherule necks7,9. Thus, the structure of the spikes demonstrates the coupling between oligomerization, allosteric activation of the capping domain, membrane binding and membrane bending in the assembly of the nsP1 complex. This is likely to be conserved among alpha-like viruses26,27.

Structural basis of the capping reaction

In contrast to conventional capping systems (which use sequential capping and methylation reactions to form the cap-1 structure (m7GpppRNA)), alphavirus nsP1 first methylates the GTP, covalently binds m7GMP and finally transfers m7GMP to a 5′-end diphosphate viral RNA acceptor13. Our structure shows that this is achieved by a modified SAM-dependent MTase-like domain. Superposition of the nsP1 capping domain with E. cuniculi MTase bound to S-adenysol-l-homocysteine (SAH) and GTP19 enabled us to identify residues, defining their binding sites for the guanine-N7 methylation reaction (Fig. 3, Extended Data Fig. 9a, b). The GTP-binding site is formed by residues close to the base moiety-binding position (A40, D152, Y285, Y154 and Y248) and positively charged and polar residues near the triphosphate position (K49, H45, R41, R70, R71, S44, Q151 and D152) (Fig. 3a, Extended Data Figs. 3g, 9a). The SAM-binding pocket is defined by residues shared with the GTP-binding site (R70, R71, Q151 and D152) and others that are exclusive for SAM binding (V153, G65, R92, E88, D89 and P83) (Fig. 3b, Extended Data Figs. 3h, 9b).

Fig. 3: Structure of the capping active site of nsP1.
figure 3

a, Cartoon representation of the nsP1 GTP-binding-site cavity. The GTP moieties (represented as lines) are superposed in two positions, from the structures of E. cuniculi in complex with m7GpppG (PDB code 2RI2) alone (red sticks), and m7GpppG and SAH (PDB code 1RI1) (cyan sticks). Only the GTP moiety of the cap structure that was visible in these PDB entries is shown. The amino acid residues that define the cavity and form putative contacts to the GTP are shown as sticks and labelled. The secondary structures that shape the cavity are also labelled. The catalytic histidine H37 is shown as sticks and highlighted by a transparent sphere envelope. GTP(SAH+) label indicates the GTP position in the structure that includes SAH. GTP label indicates the GTP position in the structure with only m7GpppG. b, Cartoon representation of the nsP1 SAM-binding-site cavity. The cavity is shown as in a, with the carbon backbone of the SAH molecule coloured as green lines, from superposition of E. cuniculi MTase in complex with m7GpppG and SAH (PDB code 1RI1). The Zn atom is represented as a shadow sphere next to P83 and behind αB (Extended Data Fig. 9).

The nsP1–m7GMP intermediate is formed by a covalent bond between H37 and the α-phosphate of the m7GTP14 (Extended Data Fig. 9a, c). In our structure, H37 is close to the GTP-binding site but about 6–9 Å away from the superposed GTP α-phosphate position. Thus, after GTP methylation, a relocation of m7GTP in the binding pocket or a conformational change must occur to permit the formation of the covalent intermediate nsP1–m7GTP. The diphosphate viral RNA (the acceptor of the m7GMP moiety) could access the active site through a positively charged path that runs from the cone-shaped chamber of the ring to complete the cap transfer (Extended Data Fig. 9d). However, the cap transfer requires the generation of the diphosphate viral RNA by nsP2. Thus, coupling of the activities of the two proteins is required for capping viral RNA during infection. Further biochemical and structural studies will be necessary to understand the nsP1–nsP2 coupling and the precise role of the identified nsP1 residues in specificity and coordination for carrying out the three steps of the viral RNA capping reaction.


To our knowledge, our structure of the nsP1 complex is the first to define the structural basis of membrane binding for the replication complex of a positive-sense single-stranded RNA virus. The nsP1 pore reveals the architecture of replication complexes in the membranes, and controls the transit of molecules that enter and exit spherules that contain the double-stranded RNA viral genome, acting as a kind of viral nuclear pore complex. This markedly improves our view of the mechanisms of action of replication complexes (Extended Data Fig. 10). Moreover, the high oligomerization order and the unusual spike-mediated interaction with the membrane makes nsP1 the first example of a new class of membrane-bending, monotopic membrane proteins28,29. This class appears to be common to several positive-sense single-stranded RNA viruses. Our results thus open up avenues for the study of these viral replication complexes in the context of their membrane association.


No statistical methods were used to predetermine sample size. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.

Expression and purification of CHIKV nsP1 from insect cells

nsP1 was expressed in Hi5 cells (Thermo Fisher). The coding sequence of nsP1 (corresponding to residues 1–535 of the non-structural polyprotein sequence from the S27 African prototype, UniProt identifier: Q8JUX6) was synthesized by Gen9 as a codon-optimized gene and subcloned into a pFastBac vector (Thermo Fisher) with a C-terminal heptahistidine tag. Recombinant bacmids and viruses were produced in YFP-DH10 Bac cells16 and Sf21 cells (Thermo Fisher), respectively, according to the manufacturer’s protocols. For protein expression, Hi5 cells in suspension at a density of 0.5–1.0 × 106 ml−1 were infected with baculoviruses at 2% volume of the culture. Cells were collected at 2–3 days after cell arrest, when nsP1 expression was ascertained to be highest by western blotting.

Cells were resuspended in buffer A (35 mM Tris pH 7.6, 200 mM NaCl, 2 mM TCEP, 5% glycerol and10 mM imidazole) supplemented with 1 mM PMSF, 10 μg × ml−1 DNase, 2 μg × ml−1 RNase and 2 mM MgSO4, and lysed by sonication. Lysates were centrifuged at 15,000g to remove unbroken cells and debris, and then again at 100,000g to pellet the membrane fraction. Membranes were resuspended at 50 mg × ml−1 in buffer A containing 1 mM PMSF using a hand-held homogenizer, and then incubated with 1% fos-choline 12 detergent (Anatrace) for 2 h at 4 °C with gentle agitation for solubilization. Solubilized membranes were then recentrifuged at 100,000g, and supernatants containing nsP1 incubated with Ni-NTA resin (GE Healthcare) in batch at 4 °C for 30 min. Resin was washed with 10 column volumes of buffer A containing 40 mM imidazole and 0.13% fos-choline 12, and the protein eluted over 1.5 column volumes with buffer A containing 250 mM imidazole and 0.13% fos-choline. Elution fractions containing the protein were pooled and concentrated (Amicon 100k MW cut-off) for size-exclusion chromatography. Samples were applied to a Superose 6 10/30 column (GE Healthcare) pre-equilibrated with 25 mM tris pH 7.6, 200 mM NaCl, 2 mM TCEP and 0.13% fos-choline 12. Fractions were analysed by SDS–PAGE to assess purity, and western blotting with an nsP1 antibody and matrix-assisted laser desorption/ionization–time of flight (MALDI–TOF) mass spectrometry to confirm the presence of nsP1.

Expression and purification of CHIKV nsP1 from E. coli

The nsP1 gene was cloned into a pET28b vector in frame with a C-terminal heptahistidine tag for expression in E. coli Rosetta pLysS (DE3) cells (Novagen). Cells were grown in 2TY medium at 37 °C until reaching an optical density of 0.8 at 600 nm, when isopropyl-β-d-thiogalactopyranoside was added at 0.5 mM to induce expression at 18 °C overnight. Cells were collected by centrifugation at 4,000g and frozen until use. Cell pellets were resuspended in 35 mM Tris HCl pH 7.6, 0.3 M NaCl, 2 mM TCEP, 5% glycerol and 10 mM imidazole (buffer A) supplemented with 1 mM PMSF, 10 μg × ml−1 DNase, 2 μg × ml−1 RNase and 2 mM MgSO4 before lysis by sonication. Following centrifugation at 15,000g, the soluble fraction was applied to a 5-ml HisTrap crude FF column (GE Healthcare) pre-equilibrated in buffer A. The column was washed with buffer A containing 40 mM imidazole, and then eluted in buffer A containing 250 mM imidazole. Elution fractions containing nsP1 were concentrated (Amicon 30k MW cut-off) before application on a S200 10/30 size-exclusion column pre-equilibrated in 25 mM Tris-HCl pH 7.6, 0.2 M NaCl and 2 mM TCEP. Fractions were analysed by SDS–PAGE, MALDI–TOF and western blotting with an anti-nsP1 antibody.

Electron microscopy

The central fraction of the peak corresponding to nsP1 oligomers was selected for analysis by electron microscopy. For negative staining, 6 μl of the protein at a concentration of 60 μg × ml−1 was applied to Cu 300 mesh carbon grids (Agar Scientific), previously glow-discharged for 1 min at 15 mA using a PELCO EasiGlow cleaning system (Ted Pella). Grids were blotted, washed twice with water and stained with 6 μl of 2% (w/v) uranyl acetate before blotting and air drying. Images were acquired using a FEI Tecnai G2 Spirit microscope operating at 120 kV and equipped with a Veleta CCD camera (Olympus).

For cryo-EM, protein was diluted to 0.25 mg × ml−1 in size-exclusion buffer lacking detergent to reduce the concentration of free micelles applied to grids (to 0.5× the critical micelle concentration). Four μl of protein was applied to glow-discharged carbon-coated Cu/Rh 300 mesh Quantifoil R2/2 holey grids (Quantifoil), before manual blotting and plunge-freezing in liquid ethane using a Leica EM CPC cryo-fixation unit. Data were collected with a Titan Krios microscope operated at 300kV (ESRF-CM01)30 equipped with a post column LS/967 energy filter (Gatan) (slit width 20 eV) and K2 Summit direct electron detector (Gatan) in counting mode. Using EPU automated software (FEI), 5,626 movies were collected at a nominal magnification of 165,000× (corresponding to a sampling rate of 0.827 Å per pixel) across a defocus range of 1 to 2.5 μm. Each movie was recorded with a dose rate of 5.27 e per pixel per s for an exposure time of 5 s distributed over 40 frames, yielding a total accumulated dose of 38.5 e per Å2.

Cryo-EM data processing

All data were processed using RELION 3.031. Movies were corrected for drift using MotionCorr2 with dose-weighting32 and contrast transfer function (CTF) determination from the aligned non-dose-weighted micrographs was performed with CTFFind433. All frames of the movies were included for motion correction and CTF estimation. Poor-quality micrographs were discarded, yielding a subset of 5,148 images. Around 1,000 particles were selected from 100 images and subjected to 2D reference-free alignment to yield templates for the autopicking procedure of RELION. Following autopicking across the curated micrographs, 263,415 particles were extracted from the dose-weighted micrographs and subjected to 2D classification and particle sorting for removal of bad particles. Inspection of the 2D classes indicated that both single and double rings of nsP1 were present in the data, and data processing beyond this point was optimized to separate these species.

The best 2D classes (corresponding to 180,921 particles) were selected for 3D classification, using a single ring model as a template that had been generated ab initio from 2D classes obtained from a previous dataset collected from the same grid on a TALOS Arctica microscope (CNB-CIB cryo-EM facility). The model was low-pass-filtered to 60 Å, and 3D classification was performed with coarse alignment sampling (7.5°) using a circular mask of 290 Å and 3–5 classes. Ab initio model generation and all classifications were performed without imposing symmetry. Three-dimensional classification consistently yielded one major class corresponding to single rings (105,825 particles), and one class corresponding to double rings (47,338 particles). A further round of 3D classification was performed with the separated single-ring and double-ring classes to improve homogeneity, until further classification yielded no improvement. Three-dimensional classification performed with finer angular sampling did not indicate that there were conformational differences between individual protomers. A 3D reconstruction of particles from the best 3D class corresponding to single rings (94,018 particles) was performed in C1 using 3D auto-refine to a resolution of 4.3 Å (4.0 Å after map sharpening), and inspection of maps did not reveal conformational differences between protomers. The particles were thus 3D auto-refined imposing C12 symmetry to a resolution of 3.7 Å without masking, and 3.0 Å with masking, as estimated using the gold Fourier shell correlation criterion at a threshold of 0.143. Resolution improved to 3.4 Å and 2.9 Å following map sharpening, and further to 2.6 Å following a single round of particle polishing and per-particle CTF correction. Masks to perform map sharpening or refinements were generated through low-pass-filtering of the model to 15 Å and extending the map by 8 pixels and adding 6 pixels of a soft edge.

Particles from the best 3D class corresponding to double rings (47,338 particles) were auto-refined as described for single rings. Reconstructions in C1 reached a resolution of 7.1 Å (6.8 Å after map sharpening), and reconstructions performed imposing D12 symmetry reached a resolution of 3.87 Å without masking and 3.1 Å with masking. A single round of particle polishing and per-particle CTF correction finally improved the resolution to 2.9 Å. Masks were generated as outlined for single rings. Local resolution in the 3D structures was estimated using the localres implementation in RELION.

Model building

De novo model building of nsP1 was performed though fitting the protein sequence into C12- and D12-sharpened maps in Coot34. The quality of the maps allowed for unambiguous assignment of all residues with the exception of the two N-terminal residues (MG), residues 365–375, residues 451–457 and C-terminal residues 473–535, for which no density was observed. Density for residues 415–420 in MBO loop 2 within the micelle and C-terminal helix 458–472 is weaker, consistent with variations in local resolution estimates. Iterative cycles of model building and real-space refinement were performed with Coot and Phenix35 with NCS to improve the model. Validation of the model was performed using MolProbity36, and through cross-correlation analysis of the model and maps in Phenix. Pocket dimensions in the structure were calculated using POCASA37. Figures were generated using Chimera38 or PyMOL.

Sucrose gradient membrane flotation assays

Discontinuous sucrose gradients and flotation assays were adapted from a previously published method15. Small-scale cultures (10 ml) of E. coli or Hi5 cells expressing nsP1 were prepared as outlined in ‘Expression and purification of CHIKV nsP1 from insect cells’ and ‘Expression and purification of CHIKV nsP1 from E. coli’. Cells were resuspended in 2 ml of lysis buffer (35 mM Tris HCl, pH 7.6, 0.3 M NaCl and 2 mM TCEP) supplemented with 1 mM PMSF, 2 μg × ml−1 RNase and DNase and lysed by sonication. Following centrifugation at 15,000g, the soluble fraction was diluted with a 67% sucrose solution to yield a final concentration of 60%. The gradient layers (bottom to top) were as follows; 2 ml of the 67% solution, 2.5 ml of the sample, 5 ml of a 50% sucrose solution, followed by 2.5 ml of a 10% sucrose solution. All sucrose solutions were prepared in lysis buffer. Samples were centrifuged at 100,000g for 18 h in a SW40 Ti swinging-bucket rotor (Beckman Coulter) to allow the membrane fractions to float to the 10–50% interface. One-ml fractions were taken from the top of the gradient and analysed by SDS–PAGE.

Activity assays and western blotting

Five μM nsP1 was incubated with 50 μM m7GTP and 100 μM SAH or 50 μM GTP and 100 μM SAM at 30 °C for 1 h in gel filtration buffer in 20-μl reaction volumes. Two μl of the reactions were applied to 10% SDS–PAGE gels for electrophoresis at 180 mV for 1 h. Samples were transferred to nitrocellulose membranes via electroblotting at 320 mA for 1 h. Following blocking of membranes with 5% milk in TBST buffer (50 mM tris-HCl pH 7.5, 150 mM NaCl and 0.05% Tween-20) for 1 h at room temperature, membranes were incubated overnight with a primary antibody raised against nsP1 (provided by A. Merits) or an anti-m3G/m7G antibody (Synaptic Systems). Both antibodies were used at a 1/10,000 dilution in TBST buffer containing 0.5% milk. Membranes were subjected to four 10-min wash steps in TBST, before incubation with a secondary-HRP conjugated antibody for 1 h, repetition of the wash steps and revelation with Amersham Start ECL reagent (GE Healthcare). Chemiluminescence was detected with a Kodak DS Digital Science Image Station.

Detergent screening for solubilization of nsP1 from membranes

Membranes were prepared from Hi5 cells expressing nsP1 as outlined in ‘Expression and purification of CHIKV nsP1 from insect cells’. Following resuspension at 50 mg × ml−1 with a hand-held homogenizer, aliquots of the suspension were incubated with 1% detergent for 3 h at room temperature or at 4 °C overnight. Following centrifugation of samples at 100,000g for 2 h, supernatants were analysed by SDS–PAGE and assayed for guanyltransferase activity as outlined in ‘Activity assays and western blotting’. The following detergents were tested: cymal-5 (5-cyclohexyl-15-cyclohexyl-1-pentyl-β-d-maltoside), anzergent 3-10 (N-decyl-N,N-dimethyl-3-ammonio-1-propanesulfonate), DG (n-decyl-β-d-glucopyranoside), β-OG (n-octyl-β-d-glucoside), fos-choline 12 (n-dodecylphosphocholine), DM (n-decyl-β-d-maltoside), sodium dodecyl sulfate (SDS), DMNG (decyl maltose neopentyl glycol), LDAO (lauryldimethylamine-N-oxide) and DDM (n-dodecyl-β-d-maltopyranoside).

Acyl capture experiments

The palmitoylation states of purified nsP1 samples were assessed with an acyl capture kit (Badrilla). In brief, 150 μg of oligomeric or monomeric nsP1 was incubated at 60 °C for 4 h in a buffer to block free thiols before acetone precipitation. The precipitates were redissolved and incubated with a reagent designed to cleave any palmitoyl or acyl moieties, in the presence of resin that covalently bound the liberated thiols. Paired negative controls performed with a thiol protection reagent in place of the cleavage reagent yielded estimates of background binding to the resin. Bound samples were eluted from the resin through heating in the presence of Laemmli buffer. Monomeric nsP1 purified from E. coli was also included as a negative control.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.