Structural Insights into Reovirus σ1 Interactions with Two Neutralizing Antibodies

ABSTRACT

Reovirus attachment protein σ1 engages glycan receptors and junctional adhesion molecule-A (JAM-A) and is thought to undergo a conformational change during the proteolytic disassembly of virions to infectious subvirion particles (ISVPs) that accompanies cell entry. The σ1 protein is also the primary target of neutralizing antibodies. Here, we present a structural and functional characterization of two neutralizing antibodies that target σ1 of serotype 1 (T1) and serotype 3 (T3) reoviruses. The crystal structures revealed that each antibody engages its cognate σ1 protein within the head domain via epitopes distinct from the JAM-A-binding site. Surface plasmon resonance and cell-binding assays indicated that both antibodies likely interfere with JAM-A engagement by steric hindrance. To define the interplay between the carbohydrate receptor and antibody binding, we conducted hemagglutination inhibition assays using virions and ISVPs. The glycan-binding site of T1 σ1 is located in the head domain and is partly occluded by the bound Fab in the crystal structure. The T1-specific antibody inhibited hemagglutination by virions and ISVPs, probably via direct interference with glycan engagement. In contrast to T1 σ1, the carbohydrate-binding site of T3 σ1 is located in the tail domain, distal to the antibody epitope. The T3-specific antibody inhibited hemagglutination by T3 virions but not ISVPs, indicating that the antibody- and glycan-binding sites in σ1 are in closer spatial proximity on virions than on ISVPs. Our results provide direct evidence for a structural rearrangement of σ1 during virion-to-ISVP conversion and contribute new information about the mechanisms of antibody-mediated neutralization of reovirus.

IMPORTANCE Virus attachment proteins mediate binding to host cell receptors, serve critical functions in cell and tissue tropism, and are often targeted by the neutralizing antibody response. The structural investigation of antibody-antigen complexes can provide valuable information for understanding the molecular basis of virus neutralization. Studies with enveloped viruses, such as HIV and influenza virus, have helped to define sites of vulnerability and guide vaccination strategies. By comparison, less is known about antibody binding to nonenveloped viruses. Here, we structurally investigated two neutralizing antibodies that bind the attachment protein σ1 of reovirus. Furthermore, we characterized the neutralization efficiency, the binding affinity for σ1, and the effect of the antibodies on reovirus receptor engagement. Our analysis defines reovirus interactions with two neutralizing antibodies, allows us to propose a mechanism by which they block virus infection, and provides evidence for a conformational change in the σ1 protein during viral cell entry.

INTRODUCTION

Neutralizing antibodies (nAbs) are a key component of the adaptive humoral immune response to protect against viral pathogens. Among the B lymphocytes that are activated during viral infection, only a minor fraction normally produces nAbs (1, 2). In most cases, nAbs recognize exposed structures on the virus surface and act by interfering with receptor binding or cell entry events, including viral uptake into cells, capsid disassembly, or penetration of host membranes (3–8). Thus, nAbs usually target glycoproteins of enveloped viruses or attachment proteins of nonenveloped viruses. Neutralizing and nonneutralizing antibodies that bind to the virus surface also can contribute to in vivo protection by recruiting effector cells, the complement system, or the cytosolic TRIM21-proteasomal machinery (9–12). A primary aim of vaccination is the induction of a protective and broadly neutralizing immune response (13). Therefore, understanding the mechanisms of virus inhibition by nAbs can contribute to successful vaccine development.

Mammalian orthoreoviruses (reoviruses) are useful models for studies of viral pathogenesis and host immunity. Reoviruses form nonenveloped, icosahedral particles that can infect the gastrointestinal and respiratory tracts of most mammalian species but are rarely associated with disease, except in very young individuals (14, 15). The three known reovirus serotypes differ in the site of primary replication, route of spread, and central nervous system (CNS) tropism. The prototype strain type 1 Lang (T1L) spreads hematogenously, infects ependymal cells, and causes nonlethal hydrocephalus in suckling mice (16–18), while prototype strain type 3 Dearing (T3D) spreads via hematogenous and neural routes, infects neurons, and causes lethal encephalitis (16, 19, 20). Studies using reassortant reoviruses indicate that these serotype-dependent differences are linked to the S1 gene segment, which encodes outer-capsid protein σ1 (21, 22).

The trimeric σ1 protein is a filamentous molecule with a head and tail morphology that extends from each of the 12 5-fold axes of the icosahedral virion. The long, slender N-terminal tail of σ1 contains both α-helical coiled-coil and β-spiral repeat domains and anchors the protein into the capsid (23, 24). The C-terminal head of a σ1 monomer is globular, featuring an eight-stranded β-barrel (25). σ1 is the reovirus attachment protein and mediates binding to the host cell surface using two known types of receptors: cell surface carbohydrates and junctional adhesion molecule-A (JAM-A) (26–28). JAM-A is an integral component of intercellular tight junctions, barriers that form between adjacent endothelial or epithelial cells. The JAM-A-binding site is conserved among all serotypes and located at the lower, more virion-proximal part of the σ1 head domain (29). Reovirus T1L also engages the carbohydrate portion of ganglioside GM2 with its head domain (27), whereas T3D σ1 binds a range of sialylated glycans near the midpoint of its tail domain (23).

Following receptor binding, reovirus virions enter cells by endocytosis. Acid-dependent proteolysis yields infectious subvirion particles (ISVPs), the first of several disassembly intermediates. Electron micrograph reconstructions of reovirus virions and ISVPs reveal filamentous structures at the icosahedral 5-fold symmetry axes protruding from ISVPs but not from virions, suggesting a structural rearrangement of σ1 from a compact to an extended conformation during virion-to-ISVP conversion (30, 31).

The reovirus-induced neutralizing immune response is most strongly directed against the σ1 protein, which is the reovirus protein whose sequence differs the most among strains (24, 25, 32). Monoclonal antibodies (MAbs) directed against σ1 are often serotype specific, and their capacity to neutralize viral infection has been demonstrated by plaque reduction neutralization assays (33–36). The monoclonal IgG2a nAbs 5C6 and 9BG5 target the σ1 proteins from T1 and T3 reoviruses, respectively. Cross-reactivity of each of these nAbs with the noncognate σ1 protein occurs and, in most cases, the extent of the 9BG5 cross-reactivity is slightly higher than that of 5C6 (37–39). Both nAbs are highly effective in vivo at protecting neonatal mice from reovirus-induced disease (34, 37), with each nAb blocking spread without affecting primary replication. 5C6 and 9BG5 also inhibit virus-induced hemagglutination (HA) (34, 39, 40). The specific residues required for efficient neutralization by 5C6 and 9BG5 have been mapped to the σ1 head by analysis of sequence changes in neutralization-resistant viral mutants (35, 41), but the precise binding epitopes and mechanisms of neutralization by these antibodies are not known.

To gain insight into the mechanisms of reovirus neutralization, we determined the crystal structures of complexes of T1L σ1 bound to the antigen-binding fragments (Fabs) of 5C6 and T3D σ1 bound to 9BG5 Fabs. The structures show that each of the Fab binding epitopes bridges two σ1 subunits, stabilizing the trimeric structure. Furthermore, the epitopes are distinct from the JAM-A-binding sites of the two proteins. We conducted surface plasmon resonance (SPR) studies to determine the affinity of each Fab for the cognate σ1 protein and to assess whether a saturated σ1-antibody complex can bind JAM-A. We used flow cytometry to test inhibition by 5C6 and 9BG5 of virus binding to JAM-A on the surface of susceptible cells. Hemagglutination inhibition (HAI) for T1 and T3 virions and ISVPs by antibodies and Fabs was quantified to investigate the effects on glycan receptor engagement. Taken together, findings from these studies provide new insights into the recognition of σ1 by nAbs and suggest that both nAbs neutralize reovirus infection by interfering with receptor binding. As both nAbs also engage epitopes that bridge two monomers, nAb binding may additionally contribute to neutralization by preventing structural rearrangements of σ1.

RESULTS

Blockade of reovirus infectivity by 5C6 and 9BG5 MAbs and Fabs.

To verify the capacity of 5C6 and 9BG5 to block reovirus infectivity in a strain-specific manner, T1L and strain T3SA⁺ virions and ISVPs were incubated with antibodies or Fab fragments prior to testing for infectivity. L929 murine fibroblasts (L cells) were adsorbed with virus-antibody mixtures, and infectivity was detected by immunofluorescence staining and compared to that of virions or ISVPs incubated without antibody. Both 5C6 MAbs and Fabs blocked T1L virion and ISVP infectivity in a similar dose-dependent manner and inhibited the infectivity of T3SA⁺ virions only at the highest antibody concentration tested (Fig. 1A). 9BG5 MAbs neutralized T3SA⁺ and (to a lesser extent) T1L virions and ISVPs, while 9BG5 Fabs primarily decreased L cell infection by T3SA⁺ virions (Fig. 1B). Our results are consistent with the previously established strain specificity of 5C6 and demonstrate its poor cross-reactivity with T3SA⁺. The capacity of 9BG5 to interact with T1 strains has been previously described. However, neutralization was not observed in classical plaque reduction neutralization assays (33, 34). We found that 9BG5 MAbs inhibited infection of L cells by T3SA⁺ ISVPs less efficiently than these MAbs inhibited infection by T3SA⁺ virions. However, 9BG5 MAbs inhibited L cell infection by T1L virions and ISVPs to a similar extent (Fig. 1B). Similarly, 9BG5 Fabs were poor inhibitors of L cell infection by T1L virions and ISVPs as well as T3SA⁺ ISVPs, but these Fabs significantly inhibited infection by T3SA⁺ virions.

FIG 1.

Capacity of MAbs and Fabs to inhibit infection of L cells by reovirus virions and ISVPs. The relative percentage of infected L cells produced by T1L and T3SA⁺ virions and ISVPs incubated with PBS or 4-fold dilutions of 5C6 (A) or 9BG5 (B) MAbs or Fabs (at concentrations spanning from 10 to 0.04 μg/ml, indicated by the triangles) was determined. Negative (−) and positive (+) controls consisted of virions or ISVPs incubated with PBS and polyclonal reovirus antiserum, respectively. Fixed cells were scored for DAPI and reovirus antigen by indirect immunofluorescence. Results are expressed as the mean percentage of virus-infected cells relative to the number of negative-control cells plus the standard error (error bars) from four fields of view per well in duplicate wells from three independent experiments. Values that differ significantly from those for particles incubated with PBS only (−) by one-way ANOVA with Dunnett's multiple-comparison test are indicated (*, P < 0.05; **, P < 0.01).

Structure of T1L σ1 in complex with 5C6 Fabs.

To provide a structural framework for rationalizing the mechanism of neutralization, we determined the crystal structure of a stable complex of the T1L σ1 head and 5C6 Fabs. At a resolution of 3.7 Å, the interaction area is well-defined in the final electron density map, and many side chains are visible. Three Fab molecules contact the trimeric σ1 head in a nearly identical manner. The entire complex is included in the asymmetric unit. Therefore, the interactions are not linked by crystallographic symmetry, providing support for their validity. While the variable domains V_L and V_H are well ordered, the constant domains C_L and C_H of 5C6 are mobile and poorly defined in the electron density.

T1L σ1 residues 308 to 470 fold into one β-spiral repeat and the globular, C-terminal head domain (27, 29). The upper part of the σ1 head is engaged by domains V_L and V_H of the three Fab molecules (Fig. 2A to C; Table 1). Each Fab bridges two σ1 subunits, making extensive contacts that bury from the solvent an interaction surface of 727 Ų (light chain and heavy chain contributions, 171 Ų and 556 Ų, respectively). As is typical for antibodies, the contacts exclusively involve the complementarity-determining regions (CDRs) of 5C6, which are referred to as L1 to L3 for CDRs 1 to 3 of V_L, respectively, and H1 to H3 for CDRs 1 to 3 of V_H, respectively. The binding epitope is discontinuous, trough shaped, and mainly engaged by three of the six CDRs: L1, H2, and H3. L2 likely forms only a single hydrogen bond with σ1 (side chain interaction of N57^L2 and Q347^σ1), whereas L3 and H1 do not interact with the antigen.

FIG 2.

Structure of T1L σ1 in complex with 5C6 Fabs. Subunits of σ1 are colored in blue, red, and yellow. The light and heavy chains of the 5C6 Fab fragment are shown in light and dark brown, respectively. (A) Top view of the σ1 head domain surrounded by three Fab molecules. Each Fab fragment bridges two σ1 subunits. (B) Side view, showing 5C6 Fabs binding at the upper, C-terminal region of the σ1 head. (C) Close-up view of Fab-σ1 interactions. CDR loops of the light chains are highlighted in light (L1), medium (L2), and dark (L3) green, and the CDR loops of the heavy chains are colored light gray (H1), gray (H2), and black (H3). The dashed box indicates the location of the interactions shown in panel D. CDR H3 forms several hydrogen bonds with σ1 residues from four different β-strands. Two CDR H3 residues, Y102^H3 and Y104^H3, interact with σ1 residues from two subunits simultaneously.

TABLE 1.

Data collection and refinement statistics

Parametera	Value(s) forb:
	T1L σ1-5C6 Fab	T3D σ1-9BG5 Fab
Data collection statistics
Resolution range (Å)	48.99–3.7 (3.80–3.70)	49.04–3.0 (3.08–3.00)
Space group	C2	P1
Unit cell dimensions
a, b, c (Å)	223.6, 129.1, 87.9	103.2, 109.3, 131.7
α, β, γ (°)	90.0, 101.2, 90.0	103.1, 113.6, 103.5
Completeness (%)	99.9 (99.7)	98.5 (98.4)
Total no. of reflections	314,028 (24,377)	338722 (26,084)
No. of unique reflections	26,263 (1,965)	94,667 (7,020)
Rmeas (%)	29.5 (161.9)	11.6 (85.4)
CC1/2 (%)	99.1 (74.8)	99.5 (64.7)
I/σI	7.94 (1.71)	13.08 (1.89)
Wilson B (Å2)	90.72	55.87
Refinement statistics
Rwork (%)	26.10	21.20
Rfreec (%)	27.79	24.69
RMSD
Bond length (Å)	0.004	0.004
Bond angle (°)	0.813	0.661

^aI, intensity of a reflection; RMSD, root mean square deviation; CC_1/2, correlation coefficient (64).

^bThe values in parentheses represent the highest-resolution shell.

^cR_free was calculated with 5% of the data.

Although several CDRs from both V_L and V_H contact σ1, H3 makes the most extensive contacts with the protein, as judged by the buried surface area. This CDR engages one σ1 subunit, forming multiple hydrogen bonds with σ1 residues from four β-strands. A key contact involves R424^σ1, which faces the carbonyl groups of I101^H3 and G103^H3 as well as the side chain of Q417^σ1. Additionally, R424^σ1 makes cation-π interactions with Y107^H3 and Y457^σ1 (Fig. 2D). Moreover, the long H3 CDR also contacts the interface of two σ1 subunits. Residues Y102^H3 and Y104^H3 interact with σ1 residues from both subunits simultaneously. Y102^H3 forms hydrogen bonds with Q371^σ1, G448^σ1, and Q417^σ1 using its amine, hydroxyl, and carbonyl groups, respectively. G448^σ1 and D426^σ1 are contacted by the Y104^H3 side chain (Fig. 2D).

Structure of T3D σ1 in complex with 9BG5 Fabs.

To compare the strain-specific modes of antigen recognition by the two antibodies, we determined the crystal structure of the T3D σ1 head in complex with 9BG5 Fab fragments at a 3.0-Å resolution (Table 1). The constant domains C_L and C_H of most 9BG5 Fabs participate extensively in crystal packing and are, therefore, less mobile and better defined in the electron density than their counterparts from the 5C6 Fabs.

As with the T1L σ1-5C6 structure, three 9BG5 Fab molecules engage the head domain of σ1, and each Fab bridges two σ1 subunits (Fig. 3). However, the location of the interface is different, as 9BG5 engages the side of the T3D σ1 head domain, whereas 5C6 binds closer to the top of the T1L σ1 head. The total surface area buried by the 9BG5 interaction with T3D σ1 is 905 Ų, and this interface area is distributed almost equally between V_L (436 Ų) and V_H (469 Ų). Five of the six CDRs are in contact with the flat epitope. L1 forms an extended hydrogen bond network with σ1 involving eight direct interactions. Residue E419^σ1, which forms an intramolecular salt bridge with R427^σ1, is faced by L1 residues S30^L1, S31^L1, and N32^L1 (Fig. 3D). Residue N32^L1 also interacts with the carbonyl group of V392^σ1 and the guanidine group of R51^L2, which additionally forms a salt bridge with D340^σ1, located at the interface between two σ1 subunits. CDRs H1, H2, and H3 participate in intermolecular contacts with several hydrophobic interactions, a small number of hydrogen bonds, and one cation-π interaction.

FIG 3.

Structure of T3D σ1 in complex with 9BG5 Fabs. Subunits of σ1 are colored in blue, red, and yellow. The light and heavy chains of the 9BG5 Fab fragment are shown in light and dark violet, respectively. Due to the poorly defined electron density for one of the 9BG5 Fab fragments, the constant region of this fragment was modeled on the basis of better-resolved chains. (A) Top view of the σ1 head domain surrounded by three Fab molecules. Each Fab fragment bridges two σ1 subunits. (B) Side view of 9BG5 Fabs binding σ1 in the middle of the head domain. (C) Close-up view of Fab-σ1 interactions. CDR loops of the light chains are highlighted in light (L1), medium (L2), and dark (L3) green, and the CDR loops of the heavy chains are colored light gray (H1), gray (H2), and black (H3). The dashed box indicates the location of the interactions shown in panel D. The CDR L1 loop forms an extensive hydrogen bond network with one σ1 subunit. E419^σ1 is faced by the three Fab residues S30^L1, S31^L1, and N32^L1. CDR L2 residue R51^L2 forms an intermolecular salt bridge with D340^σ1 and is stabilized by a cation-π interaction with Y105^H3.

Strategies for viral escape from antibody neutralization.

The crystal structures of T1 and T3 σ1 bound to Fabs from their respective neutralizing antibodies, 5C6 and 9BG5, provide information on strategies by which reoviruses can escape neutralization. T1 reovirus variants that resist neutralization by 5C6 display two point mutations in the σ1 head, Q417K and G447S (35). Likewise, variants of T3 reovirus resistant to neutralization by the 9BG5 antibody also display two point mutations in the σ1 head, D340V and E419K (41). In the latter case, the escape mutant viruses exhibit diminished neurovirulence and altered CNS tropism in mice (42).

Of the two point mutations in T1L, Q417^σ1 is directly involved in subunit bridging with 5C6 residue Y102^H3 (Fig. 2D). Thus, the mutation Q417K likely would sterically hinder Y102^H3 and Y104^H3 interactions, and the introduced positive charge adjacent to R424^σ1 might locally alter the σ1 structure due to electrostatic repulsion. No direct interaction is observed between 5C6 and G447^σ1. Neutralization escape by the G447S mutant likely results from reduced residue flexibility. The introduction of a larger side chain may alter the local protein structure, preventing 5C6 H3 binding to σ1. For T3 reovirus, the two observed mutations, D340V and E419K, probably weaken the interaction with 9BG5 due to the direct disruption of contacts. The mutation D340V introduces an aliphatic side chain that prevents the charge-mediated interaction with 9BG5 residue R51^L2. Residue E419^σ1 is involved in the formation of several hydrogen bonds with 9BG5 (Fig. 3D). The mutation from Glu to Lys inverts the charge of the residue and likely would prevent interactions with 9BG5 residues S30^L1, S31^L1, and N32^L1.

Location of antibody- and receptor-binding sites on σ1.

The JAM-A-binding site is conserved between the reovirus serotypes and located at the lower part of the σ1 head domain (26, 29). The 9BG5 epitope is located closer to the JAM-A-binding site than is the 5C6 epitope. However, neither epitope directly overlaps the receptor-binding site (Fig. 4).

FIG 4.

Receptor- and antibody-binding sites on T1L and T3D σ1. Surface representations of σ1 (A to D) and antibody Fab fragments (B, D) and ribbon drawings of the JAM-A ectodomain (A, C) are shown. Residues of σ1 that contact receptors and antibodies within a distance cutoff of 5 Å are colored. The JAM-A-binding sites are highlighted in green and the carbohydrate-binding sites are highlighted in blue on T1L σ1 (amino acids 308 to 470; A, B) and T3D σ1 (PDB accession number 3S6X; amino acids 166 to 455 [C] and amino acids 293 to 455 [D]). The epitope of MAb 5C6 is depicted in light (light chain) and dark (heavy chain) brown on T1L σ1 (A, B), and the binding site of MAb 9BG5 is highlighted in light pink (light chain) and deep purple (heavy chain) on T3D σ1 (C, D). An open book representation of T1L σ1 and a 5C6 Fab fragment (B) and T3D σ1 and a 9BG5 Fab fragment (D) are shown. The contribution of light and heavy chains to the interface area (in square angstroms) is indicated (B, D).

Reoviruses engage cell surface glycans as coreceptors. The 5C6 epitope partly overlaps the GM2 glycan-binding site on T1L σ1. Residues Q371^σ1 and T373^σ1 are involved in interactions with GM2 and also are part of the 5C6 epitope (Fig. 5). The location of CDR H1 reduces the available space at the carbohydrate-binding site, and therefore, MAb 5C6 could hinder low-affinity engagement of the glycan receptor. The 9BG5 epitope is sufficiently distant from the carbohydrate-binding site of T3D σ1 that it would not be anticipated to interfere directly with sialic acid engagement (Fig. 4).

FIG 5.

The 5C6-binding site overlaps the glycan-binding site of T1L σ1. Secondary-structure matching (SSM) superposition (63) of T1L σ1 (light yellow) binding the GM2 glycan (PDB accession number 4GU3) and T1L σ1 (yellow) binding 5C6 Fabs. Residues Q371^σ1 and T373^σ1 interact with GM2 (cyan) as well as with CDR H3 (black) of 5C6. CDR H1 (light gray) partly covers the glycan receptor-binding site. Bound 5C6 likely would hinder GM2 binding due to reduction of the space necessary for glycan receptor engagement.

HA blockade by 5C6 and 9BG5 MAbs and Fabs.

The capacity of reovirus to produce HA is dictated by binding to sialylated glycans on the erythrocyte surface, which cross-links erythrocytes and prevents them from settling to the bottom of a well (43). To investigate the effects of 5C6 and 9BG5 on glycan receptor engagement, we performed HAI assays using human erythrocytes. Four HA units of virions or ISVPs were incubated with escalating concentrations of 5C6 or 9BG5 MAbs or Fabs prior to the addition of erythrocytes. The lowest concentration of MAbs or Fabs that inhibited HA was determined for each virus-antibody mixture (Table 2). 5C6 MAbs inhibited HA by T1L virions and ISVPs at concentrations of ≥0.31 μg/ml, whereas 5C6 Fabs inhibited HA at concentrations of ≥0.63 μg/ml. This result is not surprising, on the basis of our structural observation that 5C6 partly occludes the glycan-binding site on T1L σ1. 5C6 MAbs, which exhibited low cross-reactivity with T3SA⁺ in the infectivity blockade assay (Fig. 1A), inhibited HA by T3SA⁺ virions only at the highest concentration tested (40 μg/ml) (Table 2).

TABLE 2.

Minimal antibody/Fab concentration that inhibits HA

Antibody	Concn (μg/ml)
	T1L virions	T1L ISVPs	T3SA+ virions	T3SA+ ISVPs
5C6 MAb	0.31	0.31	40	NDa
5C6 Fab	0.63	0.63	ND	ND
9BG5 MAb	1.25	1.25	0.63	ND
9BG5 Fab	ND	ND	2.5	ND

^aND, no detectable HA inhibition at the highest antibody/Fab concentration tested (40 μg/ml).

9BG5 MAbs efficiently blocked HA by T1L virions and ISVPs at concentrations of ≥1.25 μg/ml. This result is consistent with the cross-reactivity observed for 9BG5 MAbs in the L cell infectivity blockade assay (Fig. 1B). Mapping of the 9BG5 epitope on T1L σ1 reveals several similar residues in this region of the σ1 head that likely would overlap the glycan-binding site of T1L σ1 (Fig. 6). 9BG5 MAbs and Fabs also efficiently blocked HA by T3SA⁺ virions. This finding was somewhat surprising, given that the 9BG5-binding site is distant from the glycan-binding site in T3 σ1 (Fig. 4). However, neither 9BG5 MAbs nor Fabs inhibited HA by T3SA⁺ ISVPs, particles in which σ1 has undergone a conformational change (30, 31).

FIG 6.

Conservation of the 5C6 and 9BG5 epitopes on the T1L and T3D σ1 head domains. (A) ClustalW alignment of the σ1 head domains of T1L and T3D. Residues of T3D σ1 contacting 9BG5 within 5 Å are underlined in purple, residues of T1L σ1 contacting 5C6 within 5 Å are overlined in brown, and residues involved in glycan receptor binding are overlined in blue. Residues within the antibody-binding epitopes are colored in a gradient from white to red on the basis of their similarity, with white being highly dissimilar and red being identical. (B, C) Surface representations of σ1 head domains. (B) The 9BG5 epitope of T3D σ1 is mapped on the T1L σ1 head and outlined in black. Residues in the epitope are colored as described in the legend to panel A. (C) The 5C6 epitope of T1L σ1 is mapped on the T3D σ1 head and outlined in black. Residues in the epitope are colored as described in the legend to panel A.

To investigate whether there are marked differences in antibody binding to σ1 conformers present on virions and ISVPs, we performed immunoprecipitation assays. In these experiments, 9BG5 or 5C6 MAbs as a control were coupled to magnetic protein G beads prior to incubation with equal quantities of virions or ISVPs. Bound particles were resolved by SDS-PAGE, stained with colloidal Coomassie, and quantified (Fig. 7). Nearly all of the input T1L virions and ISVPs were immunoprecipitated by 5C6 or 9BG5 MAbs, consistent with the cross-reactivity observed in previous experiments (Fig. 1 and Table 2). While 5C6 MAbs immunoprecipitated T3SA⁺ virions and ISVPs relatively poorly, 9BG5 MAbs efficiently immunoprecipitated T3SA⁺ virions and ISVPs (Fig. 7). Importantly, no significant difference in the efficiency of T3SA⁺ virion and ISVP immunoprecipitation by 9BG5 was detected. This finding suggests that the reduced capacity of 9BG5 MAbs to inhibit L cell infection and HA is not due to reduced antibody recognition of the conformer of σ1 present on T3SA⁺ ISVPs.

FIG 7.

Immunoprecipitation of virions and ISVPs by 5C6 and 9BG5 MAbs. (A, B) Colloidal Coomassie-stained SDS-polyacrylamide gels in which input or immunoprecipitated T1L (A) or T3SA⁺ (B) virions (V) or ISVPs (I) were resolved. The δ fragment results from cleavage of μ1C, which occurs during ISVP formation. Bands representing the heavy (C_H) and light (C_L) chains of the MAbs, the μ1C proteins of virions, and the δ fragments of ISVPs are indicated. (C) Quantitative comparison of immunoprecipitation efficiency. Intensities of the μ1C bands for virions and the δ bands for ISVPs were quantified by infrared imaging. Results from three independent experiments are expressed as the mean band intensity relative to the input plus the standard deviation (error bars). Values that differ from 1.0, on the basis of a one-sample t test, are indicated (**, P < 0.01). An unpaired t test comparing the relative band intensities for T3SA⁺ virions and ISVPs immunoprecipitated with 9BG5 MAbs shows no significant difference (ns).

Affinity determination and investigation of JAM-A binding to a σ1-nAb complex.

To obtain affinity and kinetic data for the interaction of σ1 with Fabs, we conducted surface plasmon resonance (SPR) experiments. We immobilized the trimeric head domains of T1L σ1 or T3D σ1 on CM5 biosensor chips and used purified, monomeric 5C6 or 9BG5 Fabs, respectively, as analytes. Each Fab bound its target σ1 with a low-nanomolar affinity following a 1:1 Langmuir interaction model (Fig. 8). This model is appropriate for the analysis, as the trimeric σ1 head domain has three identical binding sites for Fabs, and adjacent Fabs do not interfere with each other during complex formation. The Fab-σ1 interactions have similar K_D (equilibrium dissociation constant) values (2.0 ± 0.1 nM and 2.5 ± 0.2 nM for the T1L-5C6 and T3D-9BG5 interactions, respectively), but the binding kinetics differ. The association and dissociation rates for the T1L-5C6 complex are significantly lower than those for the T3D-9BG5 complex.

FIG 8.

Kinetics of Fab binding to σ1. Representative kinetic binding studies of the T1L σ1 head domain with 5C6 Fabs (A) and the T3D σ1 head domain with 9BG5 Fabs (B) using SPR. The σ1 protein was immobilized, and 2-fold dilutions of Fab fragments were injected in duplicate over the biosensor surface. Fab concentrations are indicated to the right of each sensogram and ranged from 280 nM to 8.8 nM for 5C6 (A) and 37.5 nM to 0.6 nM for 9BG5 (B). The data sets were globally fitted (red lines) to a 1:1 Langmuir binding model. (C) The corresponding χ² values are indicated. The K_D values, association constants (k_a), and dissociation constants (k_d) and their standard deviations were determined using several independent measurements.

To investigate whether JAM-A can bind to σ1 when it is engaged by neutralizing antibodies, we performed SPR measurements with the T3D σ1 head covalently immobilized on a biosensor chip (Fig. 9A). 9BG5 MAbs were injected first, and to ensure that all accessible σ1 epitopes were bound by the IgG molecules, 9BG5 Fab fragments were injected immediately afterwards. The Fab fragments did not further increase the response level, indicating surface saturation by the antibody. A second injection of 9BG5 antibodies to surface saturation was followed by injection of the JAM-A ectodomain. A significant increase in the response level occurred following injection of the JAM-A ectodomain, suggesting that soluble JAM-A binds to the T3D σ1-9BG5 MAb complex. After surface regeneration, similar amounts of JAM-A bound to σ1 alone and σ1 complexed with 9BG5. An analogous experiment with T1L σ1 and 5C6 yielded similar data, showing that soluble JAM-A can bind to the preformed T1L σ1-5C6 MAb complex (Fig. 9B).

FIG 9.

Binding of the soluble ectodomain of reovirus receptor JAM-A to T3D σ1 complexed with 9BG5 MAbs (A) or T1L σ1 complexed with 5C6 MAbs (B). Shown are referenced SPR sensograms. (A) 9BG5 antibodies were injected over a chip surface coated with an immobilized T3D σ1 head. Fabs were injected immediately afterwards to determine whether all accessible σ1 epitopes were bound by the MAbs. A second injection of 9BG5 MAbs to surface saturation was followed by injection of the JAM-A ectodomain. The response level increase (30 RU) suggests binding of soluble JAM-A to a T3D σ1-9BG5 MAb complex. The response levels of 9BG5 Fabs and JAM-A (27 RU) to σ1 alone were evaluated after surface regeneration and show that similar amounts of JAM-A can bind to σ1 alone and a σ1-MAb complex. (B) 5C6 Fabs were injected to test the response to T1L σ1 immobilized on the chip surface. After a short dissociation time, the surface was regenerated and the JAM-A ectodomain was injected to test its response to T1L σ1 (71 RU). No regeneration was necessary due to the fast dissociation. 5C6 antibodies were injected twice, and Fabs were injected immediately afterwards to determine whether all accessible σ1 epitopes were bound by the MAbs. The JAM-A ectodomain was then injected, and the response level increase (67 RU) suggested binding of soluble JAM-A to a T1L σ1-5C6 MAb complex. After surface regeneration, the response levels of JAM-A (70 RU) and 5C6 Fabs to σ1 alone indicate that similar amounts of JAM-A can bind to σ1 alone and a σ1-MAb complex.

5C6 and 9BG5 block reovirus binding to JAM-A expressed on CHO cells.

While SPR studies demonstrated the capacity of σ1 to concurrently bind soluble JAM-A and nAbs, we hypothesized that steric limitations imposed at the cell surface result in a reduced receptor-binding capacity for antibody-bound σ1. To determine the effects of 5C6 and 9BG5 on reovirus binding to JAM-A on the cell surface, 9BG5 or 5C6 MAbs or Fabs were incubated with virions prior to adsorption at 4°C with JAM-A-expressing Chinese hamster ovary (JAM-A–CHO) cells. Virus binding was detected by immunostaining and flow cytometry. 5C6 MAbs and Fabs inhibited binding of T1L virions to JAM-A–CHO cells in a dose-dependent manner and had little effect on the binding of T3SA⁺ virions (Fig. 10). 9BG5 MAbs also significantly decreased the binding of T1L virions, although they did so slightly less efficiently than did 5C6. Binding of T3SA⁺ virions was partially inhibited by 9BG5 MAbs and unaffected by 9BG5 Fabs.

FIG 10.

Antibody- and Fab-mediated inhibition of reovirus binding to JAM-A-expressing CHO cells. T1L, T3SA⁺, and T3SA⁻ virions (3 × 10⁹ particles per sample) were incubated with PBS (+) or 4-fold dilutions of 5C6 (A) or 9BG5 (B) MAbs or Fabs at concentrations spanning from 10 to 0.04 μg/ml (indicated by triangles) prior to adsorption with 10⁶ CHO cells stably expressing JAM-A. Bound virus was detected by staining with polyclonal reovirus antiserum and Alexa Fluor 488-conjugated secondary antibodies and quantified by flow cytometry. JAM-A–CHO cells that were incubated with virions and stained with secondary antibody only served as negative controls (−). Results are expressed as the mean percentage of positive (virus-bound) cells plus the standard deviation (error bars) from three independent experiments (two experiments for 0.04 μg/ml in some cases). Values that differ significantly from those for virions incubated with PBS (+) by one-way ANOVA with Dunnett's multiple-comparison test are indicated (*, P < 0.05; **, P < 0.01).

To determine the contribution of glycan engagement to JAM-A–CHO cell binding for strain T3SA⁺, we used virions of strain T3SA⁻, which is isogenic with strain T3SA⁺ except for a point mutation in the S1 gene that ablates σ1 carbohydrate binding (44). While the percentage of virus bound JAM-A–CHO cells was lower overall for T3SA⁻, 9BG5 MAbs blocked T3SA⁻ cell binding to a much greater extent than the observed blockade of T3SA⁺ binding, suggesting that glycan engagement contributes significantly to JAM-A–CHO cell binding. 9BG5 Fabs had no effect on T3SA⁺ binding but decreased T3SA⁻ binding to JAM-A–CHO cells in a dose-dependent manner. These results indicate that 9BG5 blocks the adherence of T3 virions to JAM-A, whereas it has a minimal effect on binding to glycans on the CHO cell surface. 5C6 likely inhibits the engagement of both cell surface JAM-A and glycan.

DISCUSSION

In this study, we structurally and functionally characterized the binding properties of neutralizing IgG2a antibodies 5C6 and 9BG5, which target T1 and T3 reovirus σ1, respectively. Crystal structures of 5C6 and 9BG5 Fabs in complex with their cognate antigens unequivocally identified the antibody-binding sites in the σ1 head domain (Fig. 2 and 3). In each case, the antibody epitope spans two σ1 subunits, thus recognizing and possibly stabilizing the intact trimeric structure of the σ1 head. The intersubunit epitopes are distinct from the contact regions of serotype-independent receptor JAM-A (Fig. 4). Consistent with this finding, we demonstrated that the soluble JAM-A ectodomain can bind to purified σ1 protein that is complexed with MAbs (Fig. 9). Nonetheless, reovirus attachment to JAM-A-expressing CHO cells was inhibited by cognate MAbs and Fab fragments (Fig. 10). Our results suggest that reovirus neutralization by 5C6 and 9BG5 is not mediated by direct blockade of the JAM-A receptor recognition site.

If σ1 can bind 5C6 or 9BG5 and JAM-A simultaneously, then how do 5C6 and 9BG5 effectively neutralize infection of L cells? To attach to the cell surface, reovirus is thought to employ an adhesion-strengthening mechanism in which low-affinity engagement of glycan receptors is followed by higher-affinity binding to JAM-A (44–46). Membrane-associated JAM-A is present in a variety of tissues and expressed on nonpolarized cells and the basolateral surface and tight junctions of polarized epithelial and endothelial cells. The binding site for σ1 is located on the most membrane-distal Ig-like domain of JAM-A, which extends from the cell by about 80 to 90 Å. JAM-A is engaged by residues near the bottom of the σ1 head domain, adjacent to the β-spiral repeat region of the tail. Modeling indicates that reovirus binding to JAM-A on the cell surface would bring the top of the σ1 head domain into close proximity with the cell membrane (26). When σ1 is fully ligated with 5C6 or 9BG5 antibodies, which bind near the top and middle of the head domain, respectively, the virus almost certainly could not engage JAM-A; the membrane-anchored receptor could not reach its binding site on the σ1 surface (Fig. 11). Even when bound to only one antibody molecule, it would be difficult for σ1 to bind a JAM-A molecule at a site opposite the antibody. These geometric considerations suggest that the mechanism of 5C6 and 9BG5 antibody interference with reovirus binding to JAM-A is attributable to indirect steric hindrance at the cell surface. As the 5C6 binding site is located closer to the top of the σ1 head than the 9BG5-binding site, 5C6 Fabs might have a larger effect on the steric hindrance with the cell membrane than 9BG5 Fabs, which might help explain the less pronounced effect of 9BG5 Fabs in the neutralization assay (Fig. 1). It is currently unclear whether 5C6 and 9BG5 can interfere with engagement of additional proteinaceous reovirus receptors, including Nogo receptor 1 (46).

FIG 11.

Model of the proposed mechanism of 5C6 and 9BG5 neutralization. The σ1 molecule (PDB accession number 3S6X) is colored in blue. The membrane-bound ectodomain of JAM-A (the sequence with PDB accession number 1NBQ with an additional 5 amino acids added at the C terminus using the Coot program [62]) is colored in green. Antibodies (PDB accession number 1IGT; aligned with 5C6 or 9BG5 Fabs) are colored in light/dark brown (5C6) or purple (9BG5). The cell surface is indicated by a flat surface. (A) Binding of σ1 to JAM-A on the cell surface brings the top of the σ1 head in close proximity to the cell membrane. (B, C) When bound to σ1, the bulky antibodies 5C6 (B) and 9BG5 (C) would clash with the cell surface and impede JAM-A binding (for clarity, only two antibody molecules are shown for each σ1 trimer). (B) 5C6 directly interferes with low-affinity glycan engagement by the T1L σ1 head (the structure with PDB accession number 4GU3 aligned on the structure with PDB accession number 3S6X), as its binding epitope partly occludes the glycan-binding site (orange circle). (C) The T3 glycan-binding site (orange circle) is located in the σ1 body domain, which is distant from the 9BG5 epitope in the elongated σ1 conformer on ISVPs. Steric hindrance of σ1 binding to the glycan receptor in this arrangement is not anticipated. On virions, a more compact σ1 conformer may bring the 9BG5 epitope and the glycan-binding site in close spatial proximity and permit interference with glycan receptor engagement by 9BG5.

T1 and T3 reoviruses recognize different carbohydrate coreceptors, and their respective glycan-binding sites are distinct. The reovirus HA profile is serotype dependent, but both T1 and T3 are capable of agglutinating human erythrocytes. The capacity of 5C6 and 9BG5 to inhibit HA by virions has been investigated in previous studies (33, 34, 39). Here, we tested 5C6 and 9BG5 MAbs and Fabs for HAI by virions as well as ISVPs (Table 2). 5C6 MAbs and Fabs blocked HA by T1L virions and ISVPs with a high efficiency. Our structure of 5C6 Fabs bound to T1L σ1 revealed that the 5C6 epitope is located in direct proximity to the carbohydrate-binding site and indicate that the CDR H1 loop of 5C6 would clash with glycan binding as it partly occludes the carbohydrate recognition site (Fig. 2, 4, and 5). These observations suggest that the high-affinity binding of 5C6 to σ1 directly blocks low-affinity glycan receptor engagement on human erythrocytes, regardless of the size of the antibody (MAb or Fab). Carbohydrate receptors are required for T1L reovirus binding to apical surfaces of epithelial cells, where JAM-A does not usually localize (47). When perorally administered to mice, 5C6 prevents reovirus infection of M cells overlying Peyer's patches in the intestine (36). While this hypothesis remains to be tested, our data suggest that the mechanism of 5C6 inhibition of T1L reovirus infection of these cell types is via direct blockade of σ1 binding to glycan receptors.

We found that 9BG5 MAbs and Fabs inhibit HA by T3SA⁺ virions but not by ISVPs (Table 2). Since the antibody-binding site is located in the σ1 head domain and T3 σ1 engages glycans near the midpoint of the tail, our data provide additional support for the hypothesis that σ1 adopts a more compact conformation on virions than on ISVPs. Such a compact arrangement could bring the 9BG5 epitope of the T3SA⁺ σ1 head in close spatial proximity to the glycan-binding site in the σ1 tail to interfere with glycan receptor engagement. During virion-to-ISVP conversion, σ1 likely adopts the elongated conformation observed in the crystal structure (23, 27). Here, the 9BG5 epitope is sufficiently distant from the carbohydrate-binding site (∼120 Å) that an IgG molecule would not be anticipated to sterically hinder the binding of σ1 to the glycan receptor. The conformational change in T3SA⁺ σ1 that occurs during virion-to-ISVP conversion does not, however, alter the capacity of 9BG5 MAbs to bind σ1 (Fig. 7). The differential capacity of 9BG5 to interfere with the glycan binding of T3SA⁺ virions and ISVPs can also explain the finding that 9BG5 MAbs and Fabs less efficiently inhibited infection of L cells by T3SA⁺ ISVPs than virions.

While neutralization of reovirus infectivity may be explained in large part by interference with receptor binding at the cell surface, it is possible that 5C6 and 9BG5 also impair later steps in the replication cycle. The subunit-bridging epitopes of 5C6 and 9BG5 are well placed to prevent possible structural changes in σ1 that may be important for viral uptake into cells. The 5C6 and 9BG5 epitopes each span two σ1 subunits, and thus, these epitopes are likely conformation dependent and require a fully assembled trimer. Interestingly, such a binding mode is shared by neutralizing antibodies that target other trimeric stalk-like viral attachment proteins. For example, the trimeric hemagglutinin of influenza virus contains a globular head, which mediates binding to glycan receptors, and a stem region, which contains the fusion machinery. Broadly neutralizing MAbs that target the highly conserved stem region of hemagglutinin bind across the two subunits (HA1 and HA2) of a protomer and lock the molecule in the prefusion conformation, thereby inhibiting a conformational change that is required for fusion of the viral envelope and cell membrane (48). Evidence suggests that MAb 2G4 uses a similar strategy to neutralize rotavirus, a member of the same virus family (Reoviridae) as reovirus. The spike-like rotavirus attachment protein, VP4, which contains a globular domain with a fold similar to that of the σ1 head domain, reorganizes during cell entry from a local dimer to a trimer (49). On the basis of the location of a single-residue polymorphism in a neutralization escape mutant, 2G4 is thought to bind the dimeric spike in the cleft between the heads and prevent an irreversible conformational change of the protein to a trimeric state (50).

These examples suggest that neutralizing antibodies often employ a conserved strategy in which the antibody stabilizes a particular conformation of the viral attachment protein, thereby preventing a conformational change. Binding across multiple copies of a multimeric protein is arguably the easiest way in which inhibition of such changes can be achieved. Our results show that reovirus MAbs 5C6 and 9BG5 employ the same mode of binding across subunits in the attachment protein. This strategic conservation highlights a mechanism of antibody neutralization that is conserved among enveloped and nonenveloped viruses and suggests a potential role for a σ1 conformational change in reovirus cell entry. The differential functional effects of 9BG5 on T3 reovirus virions and ISVPs provide direct evidence for a structural change of σ1 during conversion from virions to ISVPs. In future studies, neutralizing and nonneutralizing antibodies that engage different σ1 epitopes may serve as useful tools to probe the conformational rearrangements in σ1 that occur during viral entry into cells.

MATERIALS AND METHODS

Cells, viruses, and antibodies.

Spinner-adapted L929 murine fibroblasts (L cells) were grown in suspension culture in Joklik's minimum essential medium (Lonza) supplemented to contain 5% fetal bovine serum (FBS; Gibco), 2 mM l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (Invitrogen), and 25 ng/ml amphotericin B (Sigma-Aldrich). JAM-A-expressing Chinese hamster ovary (JAM-A–CHO) cells were generated by transfection of CHO cells with pcDNA3.1 plasmids encoding human JAM-A, serial passage in the presence of 0.8 mg/ml of Geneticin (Gibco), and isolation of cells expressing the highest levels of JAM-A by fluorescence-activated cell sorting. JAM-A–CHO cells were maintained in Ham's F-12 medium (Gibco) supplemented to contain 10% FBS (Gibco), 2 mM l-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (Invitrogen), and 25 ng/ml amphotericin B (Sigma-Aldrich) and grown in the presence of 0.8 mg/ml Geneticin during alternate passages.

Laboratory stocks of reovirus strains T1L, T3SA⁺, and T3SA⁻ were prepared using plasmid-based reverse genetics (51). All three viruses are isogenic except for the substitution of a T3 σ1-encoding S1 gene segment for the T1L S1 gene segment in T3SA⁺ and T3SA⁻. Virus particles were purified from infected L cells by extraction with Vertrel XF specialty fluid (DuPont) and CsCl gradient centrifugation (30). Viral titers were determined by plaque assay using L cells (37). The reovirus particle concentration was determined from the equivalence of 1 unit of the optical density at 260 nm to 2.1 × 10¹² particles (52). ISVPs were generated by treatment of virions with chymotrypsin (Sigma) (53).

Hybridomas secreting reovirus-neutralizing MAbs 5C6 (39) and 9BG5 (33) were obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA, USA). Complementarity-determining regions (CDRs) from the antibodies produced by each hybridoma were sequenced as previously described (54). Daughter cell lines producing high levels of antibody were isolated from hybridomas using in situ screening and harvesting with a ClonePix FL system (Molecular Devices, Sunnyvale, CA). The highest-producing daughter clone for each antibody was used to produce supernatant for antibody purification via protein G-affinity chromatography (GE Healthcare). Fab fragments were generated from purified monoclonal antibodies via papain cleavage and separated from the cleaved portion using protein G-agarose (Thermo Fisher).

HA and hemagglutination inhibition (HAI) assays.

Purified T1L or T3SA⁺ virions or ISVPs were distributed into 96-well U-bottom microtiter plates (Costar) at an initial concentration of 10¹¹ particles per well and serially diluted 1:2 in 50 μl phosphate-buffered saline (PBS). Human type O erythrocytes (Vanderbilt Blood Bank) were washed twice with PBS, resuspended as a 1% (vol/vol) solution in PBS, delivered to virus-containing wells at 50 μl/well, and incubated at 4°C for 4 h. One hemagglutination (HA) unit was the minimum particle number sufficient to produce HA (a shield of erythrocytes in the well). Blockade of HA activity by MAbs was assessed by incubating virions or ISVPs at four HA units per well with 1:2 serial dilutions of antibodies or Fabs, starting at 40 μg/ml, for 45 min at room temperature, before adding erythrocytes and assessing HA activity.

Assessment of reovirus infectivity by fluorescence imaging.

L cells at a density of 1 × 10⁴ per well were seeded in 96-well plates and incubated overnight at 37°C. L cells were adsorbed with serial dilutions of T1L or T3SA⁺ virions or ISVPs at room temperature for 1 h. Inocula were removed, and cells were washed with PBS and incubated in fresh medium at 37°C for 20 h. Cells were fixed with cold methanol, and reovirus proteins were detected by incubation with polyclonal reovirus antiserum at a 1:500 dilution in PBS containing 0.5% Triton X-100 at 37°C, followed by incubation with Alexa Fluor 488-labeled secondary IgG and DAPI (4′,6-diamidino-2-phenylindole). Images of four fields of view per well were captured using an ImageXpress Micro XL automated microscope imager (Molecular Devices), and total and infected cells were quantified using MetaXpress high-content image acquisition and analysis software (Molecular Devices). The effect of blockade on reovirus infectivity was tested by incubating T1L or T3SA⁺ virions or ISVPs at particle concentrations that yielded approximately 60% infected cells per well with PBS, PBS containing polyclonal reovirus antiserum diluted 1:250, or PBS containing various concentrations of 5C6 or 9BG5 MAbs or Fabs at room temperature for 1 h prior to adsorption onto L cells, fixation, staining, and quantification.

Immunoprecipitation of reovirus virions and ISVPs.

5C6 or 9BG5 MAbs (5 μg) were diluted in PBS with 0.02% Tween 20 (PBS-T20) and incubated with protein G Dynabeads (Thermo Fisher) for 15 min prior to washing in PBS-T20. T1L or T3SA⁺ virions or ISVPs (4 × 10¹⁰ per sample) were diluted in PBS-T20 and incubated with the bead-antibody complexes for 15 min prior to three washes in PBS-T20. Bound proteins were eluted from the beads by boiling in SDS sample buffer and resolved by SDS-PAGE and staining with colloidal Coomassie (Invitrogen). Destained gels were scanned using an Odyssey imager (LI-COR), and bands representing virions (μ1C) or ISVPs (δ) were quantified using Image Studio (version 5.2) software (LI-COR).

Flow cytometric analysis of reovirus binding.

T1L or T3SA⁺ virions (3 × 10⁹ per sample) were incubated with serial 4-fold dilutions, starting at 10 μg/ml, of 5C6 or 9BG5 MAbs or Fabs in PBS containing 2% FBS at room temperature for 30 min prior to adsorption with 10⁶ JAM-A–CHO cells per sample at 4°C for 30 min. Bound virions or JAM-A was detected by incubation with polyclonal reovirus antiserum or human JAM-A (hJAM-A) antibodies, followed by incubation with Alexa Fluor 488-labeled secondary IgG. For each sample, 10,000 events were analyzed for antibody binding using a BD LSR II flow cytometer (BD Biosciences). The gating controls to determine cutoffs for virus-positive cells were JAM-A–CHO cells incubated with reovirus particles and stained with secondary antibody only.

Statistical analysis.

Statistical analyses were performed using GraphPad Prism (version 4) software (GraphPad). For L cell infectivity assays, mean relative values from four fields of view in duplicate wells from three independent experiments were compared using one-way analysis of variance (ANOVA) with Dunnett's multiple-comparison test. For flow cytometry assays, mean values of the percentage of positive cells from three independent experiments were compared using one-way ANOVA with Dunnett's multiple-comparison test. For immunoprecipitation studies, mean relative band intensities from three independent experiments were compared to 1.0 using a one-sample t test, and mean values from T3SA⁺ virions and ISVPs immunoprecipitated with 9BG5 were compared to one another using an unpaired t test.

Protein production and purification.

The T1L σ1 head (amino acids 308 to 470) was expressed in Escherichia coli strain Rosetta 2 (DE3) and purified via Ni-affinity chromatography and size exclusion chromatography as described previously (27). The untagged protein contains 8 nonnative amino acids at the N terminus and was used for complex formation with 5C6 Fabs as well as for SPR measurements.

The T3D σ1 head (amino acids 294 to 455) was expressed and purified as described previously (55) with minor modifications. Expression in Escherichia coli strain BL21(DE3) cells (Novagen) was induced with 0.2 mM IPTG (isopropyl-β-d-thiogalactopyranoside) at 20°C for 16 to 20 h. Bacteria were lysed by sonication with a 40% duty cycle using a Branson model 450 sonifier and centrifuged at 15,000 × g. The protein was purified via glutathione S-transferase (GST)-affinity chromatography using a 5 ml GSTrapFF column (GE Healthcare) and eluted with 30 mM reduced glutathione, 3 mM EDTA, and 50 mM Tris (pH 8.0). The GST tag was removed by tryptic digestion at 20°C for 4 to 6 h. A PD10 desalting column (GE Healthcare) was used to change the buffer to 20 mM HEPES (pH 7.4), and the untagged protein was purified by MonoQ anion-exchange chromatography (GE Healthcare). The protein contains 2 additional nonnative amino acids at the N terminus and was used for complex formation with 9BG5 Fabs and SPR measurements. As with the T1L head, the 2 extra residues are distant from the antibody- and JAM-A-binding sites.

The extracellular region of human JAM-A (amino acids 27 to 233, comprising two Ig-like domains) was expressed in Escherichia coli and purified via affinity chromatography using a GSTrap column (GE Healthcare) and anion-exchange chromatography as described previously (56).

Complex formation.

Complexes were formed by mixing purified T1L σ1 with 5C6 Fabs or T3D σ1 with 9BG5 Fabs using a molecular ratio of one σ1 trimer to four Fabs and incubating the mixture at 4°C for 45 min. Complexes were separated from excess Fabs by size exclusion chromatography (Superdex 200 10/300) in 150 mM NaCl, 20 mM HEPES (pH 7.4). In both cases, the elution volumes from the size exclusion chromatography indicated that a stable complex with the expected size had been formed.

Crystallization and X-ray structure determination.

The T1L σ1-5C6 Fab complex was concentrated to 1 to 4 mg/ml. Crystals were grown in 9.5 to 11.5% (wt/vol) polyethylene glycol 8000 (PEG8000), 0.1 M MES (morpholineethanesulfonic acid; pH 5.5 to 6.5), and 0.2 M zinc acetate at 4°C using the hanging-drop vapor diffusion method. Crystals were flash frozen with the mother liquor containing 12% 2-methyl-2,4-pentanediol (MPD) as a cryoprotectant. Diffraction data from several crystals were collected at beamline X06DA of the Swiss Light Source (SLS; Villigen, Switzerland). The XDS program (57) was used to integrate, merge, and scale the data obtained from the two crystals. The structure was solved by molecular replacement with Phaser software (CCP4 suite) (58, 59) using the structure of the T1L σ1 head comprising amino acids 308 to 470 (PDB accession number 4GU3) and mouse IgG2a(κ) Fab 9BG5 as search models. The crystals belonged to space group C2 with one T1L σ1 trimer and three 5C6 Fabs in the asymmetric unit.

The T3D σ1-9BG5 Fab complex was concentrated to 1 to 2 mg/ml, and crystals were grown in 10 to 12% (wt/vol) PEG8000, 0.1 M Tris (pH 6.5 to 7.5), and 0.2 M magnesium chloride at 4°C using the hanging-drop vapor diffusion method. The reservoir solution was covered with silicone oil to reduce vapor diffusion due to initially fast-growing crystals. The crystals were flash frozen with the mother liquor containing 12% MPD as a cryoprotectant. Diffraction data were collected at beamline X06DA of the SLS. XDS software (57) was used to integrate and scale the diffraction data. The structure was solved by molecular replacement with Phaser software (CCP4 suite) (58, 59) using the structure of the T3D σ1 head (PDB accession number 2OJ5) and a mouse IgG1(κ) Fab (PDB accession number 1FIG) as search models. The crystals belonged to space group P1 with two T3D σ1 trimers and six 9BG5 Fab molecules in the asymmetric unit. Refinement of the two structures was performed using the programs Phenix (60) and autoBUSTER (61), and model building was done using the program Coot (62).

Surface plasmon resonance. (i) Kinetic analysis.

SPR experiments were performed using a Biacore 2000 instrument (GE Healthcare) at 25°C. Standard amine coupling chemistry (NHS/EDC kit; GE Healthcare) was used to covalently immobilize the T1L (56 kDa) or T3D (53 kDa) σ1 head to a CM5 sensor chip with immobilization levels of 37 response units (RU) and 33 RU, respectively. Deactivated flow cells served as references.

Concentration series of the Fab fragments, which served as analytes, were prepared by 2-fold dilutions in running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 0.05% polysorbate 20). For T1L σ1-binding experiments, 5C6 Fabs (47 kDa) were randomly injected in duplicate or triplicate over the biosensor surface for 500 s with a dissociation time of 1,300 s and a flow rate of 30 μl/min. For kinetic analysis of T3D σ1 and 9BG5 Fabs (47 kDa), the Fabs at each concentration were randomly injected in duplicate for 210 s, followed by a dissociation step of 450 to 600 s at a flow rate of 50 μl/min. To remove the Fab fragments from the σ1-coated surface after each cycle, 5 μl of regeneration solution (3.3 mM glycine, pH 1.7) was applied. Three independent experiments were performed in each case. Data from assays with six different 5C6 concentrations ranging from 8 to 280 nM and data from assays with seven different 9BG5 concentrations ranging from 0.6 to 37.5 nM were double referenced and fitted to a 1:1 Langmuir binding model (BIAevaluation).

(ii) hJAM-A binding to a σ1-MAb complex.

To investigate possible binding between a saturated σ1-MAb complex and hJAM-A, a Biacore 2000 system was used. The T1L or T3D σ1 head was immobilized to a CM5 sensor chip (NHS/EDC kit; GE Healthcare) with an immobilization level of 60 RU or 25 RU, respectively. In the case of T3D σ1, 9BG5 MAbs (301 nM) were injected over the biosensor surface for 180 s at a flow rate of 20 μl/min. To ensure that the T3D σ1-coupled surface was saturated by the antibody, 9BG5 Fabs (8 μM) were coinjected for 120 s. After a dissociation step, 9BG5 MAbs (301 nM) were injected to achieve surface saturation for 180 s, and the soluble ectodomain of hJAM-A (8 μM) was coinjected for 120 s. Dissociation was followed by regeneration with 3.3 mM glycine, pH 1.7. The response of 9BG5 Fabs (8 μM) or hJAM-A (8 μM) to the regenerated T3D σ1 surface was tested by injection for 120 s.

In the case of T1L σ1, the response of 5C6 Fabs (4.4 μM) to σ1 alone was first tested by an injection for 120 s at a flow rate of 20 μl/min. Dissociation was followed by regeneration with 3.3 mM glycine, pH 1.7. The response of the JAM-A ectodomain (14 μM) to σ1 alone was tested by injection for 120 s. Due to the fast off rate, no regeneration was required. 5C6 MAbs (4 μM) were injected for 600 s, followed by an additional 300 s, to achieve surface saturation. Immediately afterwards, 5C6 Fabs (4.4 μM) were injected for 120 s to determine whether all accessible binding sites of σ1 were bound by the MAb. The JAM-A ectodomain (14 μM) was injected for 120 s. Dissociation was followed by regeneration with 3.3 mM glycine, pH 1.7. Subsequently, the response of 5C6 Fabs (4.4 μM) or JAM-A (14 μM) to the regenerated T1L σ1-coupled surface was tested by injection for 120 s.

Accession number(s).

Coordinates and structure factors have been deposited with the Protein Data Bank with accession numbers 5MHS and 5MHR for the T1L-5C6 and T3D-9BG5 complexes, respectively.