Abstract
Equine rhinitis A virus (ERAV) shares many features with foot-and-mouth disease virus (FMDV) and both are classified within the genus Aphthovirus of the family Picornaviridae. ERAV is used as a surrogate for FMDV research as it does not require high-level biosecurity. In contrast to FMDV, which uses integrins as cellular receptors, the receptor for ERAV has been reported to involve the sugar moiety sialic acid. This study confirmed the importance of sialic acid for cell entry by ERAV and reports the crystal structure of ERAV particles complexed with the receptor analogue 3′-sialyllactose. The receptor is attached to the rim of a capsid pit adjacent to the major immunogenic site and distinct from the sialic acid binding site used by a related picornavirus, the cardiovirus Theiler's murine encephalitis virus. The structure of the major antigenic determinant of the virus, previously identified from antibody escape mutations, is also described as the EF loop of VP1, which forms a hairpin stretching across the capsid surface close to the icosahedral fivefold axis, neighbouring the receptor-binding site, and spanning two protomeric units.
-
↵†Present address: Institute for Animal Health, Pirbright, Woking GU24 0NF, UK.
-
A supplementary figure showing an icosahedrally averaged omit map for sialyllactose bound to ERAV is available with the online version of this paper.
INTRODUCTION
Equine rhinitis A virus (ERAV) is one of several picornaviruses that cause cold-like symptoms in horses. ERAV is genetically most closely related to foot-and-mouth disease virus (FMDV) and both viruses are classified within the genus Aphthovirus. The virus contains a positive-sense RNA genome of approximately 7700 nt with a genome organization broadly typical of picornaviruses. The genome is contained within a non-enveloped icosahedral capsid comprising 60 copies of each of the viral proteins: VP1, VP2, VP3 and VP4. In all picornaviruses, VP1–3 adopt the typical viral eight-stranded β-barrel fold with the loops connecting the strands (labelled B–I). The surface-exposed loops together with the C termini define the antigenic phenotype of the virus. We have recently determined the crystallographic structure of ERAV, which showed a broad similarity to FMDV, with the most significant differences relating to the increased length and conformation of the surface loops in VP1 (Tuthill et al., 2009). Both ERAV and FMDV dissociate into pentameric capsid subunits at low pH and this was thought to be the mechanism for genome release from the particle. However, we also recently demonstrated the existence of a low-pH-derived empty particle of ERAV that had lost its RNA, indicating that genome release and capsid dissociation may be distinct events, and the structure of a low-pH particle also provided clues to the mechanism of capsid dissociation (Tuthill et al., 2009).
ERAV and FMDV share a number of distinctive physical properties such as buoyant density, base composition and acid lability, as well as biological properties such as a broad host-cell range and the propensity for persistent infection. However, the disease caused by ERAV is quite different to that caused by FMDV and hence it is not surprising that the receptors utilized by these viruses differ.
FMDV field viruses are dependent on cell-surface integrin receptors in vitro and very probably in the infected animal. However, adaptation of FMDV to cell culture can result in the selection of virus variants that have a high affinity for heparan sulphate (Jackson et al., 1996; Sa-Carvalho et al., 1997). Heparan sulphate-adapted variants retain the ability to use integrins (Sa-Carvalho et al., 1997). Other as-yet-unidentified classes of cell-surface molecule are also thought to serve as receptors for FMDV (Baranowski et al., 2000; Zhao et al., 2003). Indeed, it appears that endocytosis of the virion by any means can result in acid-mediated uncoating to initiate productive infection (Mason et al., 1993).
In contrast, ERAV is thought to use α2,3-linked sialic acid to enter cells (Stevenson et al., 2004b). Persistent strains of Theiler's murine encephalitis virus (TMEV), a cardiovirus, also recognize sialic acid as a receptor moiety and this interaction has been studied by X-ray crystallography (Zhou et al., 2000). Sialic acid exists as N-acetyl neuraminic acid and N-glycolyl neuraminic acid forms, which are displayed at epithelial cell surfaces on the tips of oligosaccharide chains on cell-surface glycoproteins, most commonly as a single terminal residue appended to an underlying galactose or N-acetyl galactosamine by an α2,3- or α2,6-glycosidic linkage.
In all FMDVs, the exposed, mobile GH loop of VP1 is central to both the antigenicity of the virus and integrin binding, with some involvement of the C-terminal region of VP1 (Acharya et al., 1989; Fox et al., 1989; Jackson et al., 2003; Strohmaier et al., 1982). Although there has been no evidence for an equivalent receptor-binding structure in ERAV, neutralization escape mutants have defined a similar major antigenic site involving the VP1 EF loop together with the VP1 C-terminal region (Kriegshäuser et al., 2003; Stevenson et al., 2004a). Whilst ERAVs are not as divergent as the FMDV serotypes, variation has been found in the neutralizing epitopes (Varrasso et al., 2001).
Here, we confirmed the role of sialic acid as a critical component of the receptor for ERAV entry, showing the crystal structure of its interaction with the virus particle, and we discuss the bearing this has on the major antigenic site.
RESULTS
α2,3-Linked sialic acid competes with receptor for ERAV binding and reduces virus infectivity
Previous studies used neuraminidase treatment of cells to demonstrate that sialic acid is a receptor for ERAV (Stevenson et al., 2004b) and persistent strains of the cardiovirus TMEV (Shah & Lipton, 2002). In addition, infection of cells with TMEV is inhibited in the presence of sialyllactose, which presents sialic acid in the α2,3-linked configuration and binds virus in competition with cell-surface receptors (Zhou et al., 2000). We therefore investigated the ability of sialyllactose to inhibit ERAV infection of HeLa cells using an infectious centre plaque assay. The presence of sialyllactose at 10 mM resulted in at least a tenfold reduction in the infectivity of ERAV (Fig. 1a⇓), whilst having no effect on poliovirus (which uses a different receptor), thus confirming the involvement of α2,3-sialic acid as a receptor for entry of ERAV.
Evidence for sialyllactose interaction. (a) Infectious centre assay showing that sialic acid competes with the receptor for ERAV binding and reduces virus infectivity. ERAV or poliovirus type 1 (PV1) was treated with sialyllactose or mock treated and allowed to attach to HeLa cells in suspension. Cells were washed, serially diluted and plated onto confluent monolayers in order for plaques to develop. Each panel shows the plaques resulting from 10−2 and 10−3 dilutions of infected cells as indicated. Treatment with or without sialyllactose is indicated on the left. The effect of sialyllactose treatment (top row) was evident as a decrease in infectious virus (number of plaques) relative to untreated cells (bottom row). (b) Electron density (magenta) for sialyllactose (drawn as sticks coloured by atom type) as bound to ERAV observed in a Fo(soak) − Fc(native) averaged map at 4 Å resolution.
Crystallographic analysis
Density difference maps derived from crystals irradiated in the presence or absence of sialyllactose [Fo(soak)−Fo(native)] at 4 Å resolution, subjected to one cycle of 30-fold averaging, clearly showed the presence of bound sugar (Fig. 1b⇑). There was a recognizable density for the sialic acid and galactose moieties, whilst the glucose moiety appeared to be disordered. The |2Fo(soak)−Fc(native)| map indicated that the sugar density was no more than 50 % occupied (compared with the protein). A molecule of sialyllactose was modelled into the electron density map using the side-chain density of the sialic acid for orientation and the complex refined using xplor maintaining strict non-crystallographic symmetry (NCS) constraints (and keeping the occupancy fixed at 0.5). No water molecules were included in the model. An omit map [Fo(soak)−Fc, after one cycle of averaging; see Supplementary Fig. S1, available in JGV Online] clearly reaffirmed the sugar density and showed the model to be constrained by the density for at least the sialic acid moiety. The final R-factor was 24.7 % with 0.02 Å root mean square deviation (r.m.s.d.) (bond lengths). Although the crystals were grown at low pH, a comparison with the native structure showed no significant changes in the vicinity of the receptor-binding site (Tuthill et al., 2009).
Attachment site
The oligosaccharide bound in a shallow groove of biological protomer ‘A’, which followed the outer edge of the EF loop from VP1 of an adjacent clockwise related protomer as it lay across protomer ‘A’. The groove was blocked off by Val-135 and Ala-136 at the northern rim of the large surface depression (pit) (Fig. 2a, b⇓). The sialic acid was positioned adjacent to a block of positive charge contributed by Arg-129 and Gln-120 of VP1 (Fig. 2c⇓). There was no close correspondence between this site and the location of the heparin-binding site in FMDV or the sialic acid-binding site in TMEV, which was located in a positively charged depression formed by puff B of VP2, some 15 Å from the pit.
Receptor binding. (a) Surface representations of the ERAV and FMDV (PDB 1FOD) viral capsids coloured by radial height (with the same colour scheme for both particles) to illuminate surface features with bound sialyllactose and heparin (purple), respectively, and the ligand-binding sites on the respective protomers. (b) Viral protomers depicted as ribbons (De Lano, 2002): VP1, blue; VP2, green; VP3, red; VP4, yellow. In ERAV (left), the EF loop from the adjacent protomer, which forms part of the binding site for siallylactose, is shown in cyan. In FMDV (right), the GH loop with the RGD recognition motif is shown in cyan – residues implicated in antigenic sites 1 and 5 (FMDV) and site 1 (ERAV) are shown in purple. The bound receptors are coloured yellow. (c) On the left is shown a surface representation of ERAV with the proteins colour coded as in (b). Siallylactose is drawn as sticks coloured by atom (N, blue; O, red).The interacting side chains Gln-65, Arg-129 and Gln-120 are represented in a similar fashion, as is Glu-122 (mutation that is responsible for increased resistance to neutralization). The inset shows an N-acetyl neuraminic acid ring structure. To the right is a ligplot (Wallace et al., 1995) representation of the ERAV–siallylactose complex. Ligand bonds are drawn with solid purple lines and non-ligand bonds with solid black lines. Hydrogen bonds are shown as dashed light-blue lines (for clarity, some hydrogen bonds have been omitted). Non-ligand residues involved in hydrophobic contacts are shown as orange-fringed semi-circles and the corresponding atoms in the hydrophobic contact with a similar fringe. The ligand labels are colour-coded as in (b) with the EF loop residues in cyan. Hydrogen bond distances are shown in cyan.
Sialyllactose receptor conformation
At 4.0 Å resolution, it is difficult to determine sugar conformations precisely, but the pyranose ring of the sialic acid appeared to be the α-anomer in a chair conformation (carboxylate in the axial position). A detailed view of the receptor binding site is given in Fig. 2(c)⇑. The majority of the interactions were with the EF loop of the clockwise related biological protomer. Key interactions appeared to be contributed by Arg-129 of VP1 (in the EF loop) and the carboxylate of the sialic acid. Such interactions are common to many complexes with sialyllactose, e.g. in murine polyomavirus (Stehle & Harrison, 1996) and influenza B neuraminidase (sialic acid complex) where the carboxylate interacts with three arginines (Burmeister et al., 1992); however, it is not seen in the TMEV–sialyllactose complex (Zhou et al., 2000). The carbonyl oxygen of the N-acetyl group was stabilized by a weak hydrogen bond with the main chain amide of Gln-65 – this was the only interaction with residues from the biological protomer to which the sialic acid was adjacent apart from that mediated by a bridging water molecule between Sia O10 and VP1 63 CG2. The acetyl group also made a hydrophobic interaction with Trp-119 and the amino nitrogen made a hydrogen bond with Ala-118 O. The glycerol group was held extended by a hydrogen bond between the 9-hydroxyl group and the carbonyl oxygen of Gln-120, and there was also an interaction between the 8-hydroxyl group and the backbone amide of Gln-120. The 7-hydroxyl group, however, made no interactions. The galactose ring appeared to adopt a chair conformation. Neither the galactose nor the glucose made interactions with the capsid, hence their increased flexibility and our inability to clearly visualize the glucose moiety [mean B-factors: sialic acid (Sia) 18 Å2, galactose (Gal) 48 Å2, glucose (Glu) 72 Å2].
The residues that act as receptor ligands appear to be conserved across all strains of ERAV (Varrasso et al., 2001). These residues have fairly low B-factors indicating that the site of interaction is quite rigid, and indeed there was minimal movement between the apo structure and that with bound sialyllactose. Superimposition of the ERAV-bound sialic acid with the sialic acid of the polyomavirus–sialyllactose complex [Protein Data Bank (PDB) 1SID; Stehle & Harrison, 1996] gave an r.m.s.d. of 1.0 Å, with strong correspondence both in conformation and ligands.
Antigenic surface
Antigenic sites in picornaviruses have been identified from the locations of amino acid substitutions conferring resistance to neutralizing monoclonal antibodies. Although this method identifies only a subset of antibody–contact residues, it has established a correlation between antigenicity and the most accessible surface loops. Thus, the immunodominance of VP1 is attributable to it being the most surface-accessible capsid protein.
Independently isolated ERAV escape mutants selected against a panel of four neutralizing monoclonal antibodies showed extensive cross-resistance indicating that they belong to the same major immunogenic site although widely separated in sequence (Kriegshäuser et al., 2003). The selected mutations were located in the EF loop of VP1 at Lys-114 or in the C terminus of VP1 at residues Pro-240 and Thr-241. The EF loop of VP1, a finger-like projection of antiparallel β-sheet, traverses the surface of the particle in an anticlockwise direction around the fivefold axis approaching the VP1 C terminus of the adjacent protomer. This structure is well ordered in the electron density map, implying that it is fairly rigid. Thus, the major neutralizing immunogenic site is formed by amino acid residues from two adjacent VP1 subunits, as in FMDV (Acharya et al., 1989). This site is located at the inter-protomer boundary on the periphery of the fivefold plateau and the Cα atoms of residues Lys-114 and Pro-240 approach to within 6 Å (Fig. 3⇓), certainly close enough to constitute a single immunogenic site. Another residue identified as responsible for increased resistance to neutralization and probably part of a conformational epitope, Glu-122 (Stevenson et al., 2004a), is also in the EF loop but some 24 Å from Lys-114 (Fig. 2c⇑). The EF loop is one of the regions where most amino acid variation occurs between isolates (Varrasso et al., 2001). Apart from the last four residues and an insertion of two residues in ERAV, the VP1 C termini of FMDV and ERAV are topologically very similar and, as both are implicated in major immunogenic sites, the relative locations of these sites are likely to be similar.
Antigenicity. A cartoon trace of two biological protomers of ERAV with the proteins (VP1–3) colour coded by the standard convention (see Fig. 2) but brighter for one protomer. The VP1 EF loop from the adjacent (dull) protomer is shown in cyan traversing the inter-protomer boundary to come into proximity to the C terminus of VP1 from its neighbour. The VP1 residues Pro-240 and Lys-114, implicated as forming a single immunogenic site although widely separated in sequence (Kriegshäuser et al., 2003), are highlighted by drawing the residues in a space-filling representation.
Non-neutralizing B-cell epitopes are located at the N and C termini, the βE–βF and βG–βH loops of VP1, and at the N termini of VP2 and VP3 (Li et al., 2005; Stevenson et al., 2003). Of these, the three N termini are internal in the mature particle unless they become transiently exposed during capsid breathing.
DISCUSSION
The major immunogenic site of ERAV is conformational, encompassing the EF loop (residue 114) and C terminus of VP1 (residues 240 and 241) with each site formed by amino acid residues from two adjacent VP1 molecules (Kriegshäuser et al., 2003; Stevenson et al., 2004a). This site occupies a radically different position on the virus surface compared with the visualized major immunogenic site in FMDV (the GH loop) (PDB 1FOD; Berman et al., 2000) and, in contrast to the flexible FMDV GH loop, the EF loop of ERAV is well ordered. Both the FMDV GH loop and the ERAV EF loop dominate the accessible surface of the particle, both are important for receptor recognition (sialic acid for ERAV and integrin for FMDV) and both possess internal secondary structure – a β-strand and α-helix in FMDV and an antiparallel β-hairpin in ERAV. This leads to the dilemma of balancing receptor conservation against antigenic variation. There is a notable difference, however: whilst flexibility of the GH loop of FMDV appears to be a requirement for recognition of the cell-surface integrin receptor, in ERAV receptor engagement involves recognition of a rigid loop by sialic acid. FMDV, in common with a number of viruses, can, especially in tissue culture, adapt to bind sulphated sugars, probably as an initial receptor. For example, in FMDV strain O1BFS, residue 56 of VP3 switches from a histidine in field viruses to an arginine (which is a key ligand for heparin; Fry et al., 1999, 2005), resulting in a partial loss of antigenicity. Whilst ERAV strain NM11 has been adapted to growth in cell culture, the residues that interact with the sialic acid are conserved across ERAVs (Varrasso et al., 2001), so there is no suggestion that this binding is a result of adaptation to growth in tissue culture. Furthermore, the low pH of crystallization does not appear to alter the external capsid structure (Tuthill et al., 2009) and is unlikely to affect sialic acid binding.
Members of the genus Cardiovirus, such as TMEV, to which ERAV is quite closely related, also bind sialic acid (Alexander & Dimock, 2002). The residues involved in forming the sialic acid-binding site are not conserved between these viruses but there is some topological similarity. Indeed, there are marked similarities in receptor conformation and the nature of the sugar atoms interacting with the virus between ERAV, polyomavirus (Stehle & Harrison, 1996) and TMEV (Zhou et al., 2000): only the sialic acid moiety is attached to the capsids, with the rest of the oligosaccharide extending away from the surface. Thus, there is little shape complementarity and little buried surface area, with minimal changes in the capsid on binding, suggesting that the binding site is quite rigid. The nature of the glycosidic linkages is probably not a determinant of specificity in this case but rather the position of particular hydroxyl groups. The liganding residues brought together by the three-dimensional folding of the protein are important (Fig. 2b, c⇑), with a mix of electrostatic and van der Waals interactions and bridging water molecules defining a specific, low-affinity binding site, features that seem to be characteristic of sugar receptors.
METHODS
3′-Sialyllactose.
3′-Sialyllactose was used to present the sialic acid molecule in the α-configuration, which is mostly observed in the natural glycoconjugates, and was obtained from Dextra Laboratories and Sigma-Aldrich.
Growth and purification of virus.
ERAV strain NM11 was grown and purified as described previously (Tuthill et al., 2009). Infected cell lysates were clarified by low-speed centrifugation, the supernatant was precipitated with ammonium sulphate and the virus was purified by sedimentation through two sucrose density gradients. Purified virus was quantified by absorbance at 260 nm, where an A260 value of 7.7 was equivalent to 1 mg ml−1, and purity was confirmed by SDS-PAGE and Coomassie staining. Poliovirus was grown and purified as described previously (Tuthill et al., 2009). Briefly, infected cell lysates were clarified by low-speed centrifugation and the virus was purified by banding on caesium chloride density gradients.
Infectious centre plaque assay for receptor competition.
Sialyllactose was dissolved in PBS and added to purified virus to a final concentration of 10 mM and incubated for 30 min at room temperature. Subconfluent HeLa Ohio cells were removed from culture plastic ware by treatment with 10 mM EDTA in PBS and washed twice in PBS. Cell pellets (approx. 105 cells) were resuspended with the mixture of virus and sialyllactose and mixed gently for 30 min at room temperature to allow virus attachment. Cells were pelleted and washed with PBS twice to remove unattached virus, resuspended in pre-warmed culture medium (Dulbecco's modified Eagle's medium with 2 % fetal calf serum) and incubated at 37 °C for 1 h to allow infection to proceed. Infected cells were diluted in a tenfold series, seeded onto confluent cell monolayers in six-well plates and allowed to settle for 4 h. The medium was replaced with 0.6 % agarose overlay in medium. After 3–4 days, the monolayers were fixed and stained with 4 % (v/v) formaldehyde, 5 % (v/v) ethanol, 0.5 % (w/v) crystal violet in PBS.
Complex structure determination.
Crystals were grown from 1 M diammonium hydrogen citrate, 0.1 M sodium acetate (pH 4.6) (Tuthill et al., 2009). These crystals were thicker and easier to manipulate than crystals obtained at pH 7.0. They were thus chosen for these experiments, together with the knowledge that the native and low-pH structures showed no significant differences in their external capsid features. The crystals were stabilized with 1.8 M diammonium hydrogen citrate and then soaked with sialyllactose at a concentration of 50 mM for 2 h before mounting in quartz capillaries. Diffraction images were collected at room temperature as 0. ° oscillations on a CCD detector at the European Synchrotron Radiation Facility, station BM14, at λ=0.87 Å (Table 1⇓) and processed using denzo and scalepack (Otwinoski, 1993). The crystals soaked were not above 0.07 mm in maximum dimension and thus the microdiffractometer beam was collimated to 30 μm. The 4.0 Å resolution data were scaled to those for the native virus and combined with phases from the latter for the calculation of a |Fo(soak)−Fo(native)| difference electron density map using the phases for the native virus. Maps with |Fo(soak)−Fc(native)| and |2Fo(soak)−Fc(native)| coefficients were also calculated. These maps were 30-fold averaged using gap (D. I. Stuart and Jonathan Grimes, unpublished data). The electron density was commensurate with the glucose group being disordered. A molecule of sialyllactose was built into the electron density map. Having verified that there were no significant differences between the NCS-related copies, the model was refined by iterative positional and B-factor refinement (xplor) (Brunger, 1992) using NCS constraints and appropriate stereochemical parameters for sialyllactose, which was judged to be at half occupancy (Table 1⇓).
X-ray data collection and structure refinement statistics
na, Not applicable.
Acknowledgments
We thank Martin Walsh and staff at the UK MAD beamline BM14 for help with data collection, Max Crispin for advice on sugar structure and the Medical Research Council UK for financial support. Coordinates and structure factors have been deposited in the PDB (codes 2xbo and r2xbosf, respectively).