Structural studies on antibody recognition and neutralization of viruses

Abstract

The purpose of this brief review is to highlight how structural information can elucidate antibody recognition and neutralization of viruses. Studies on human rhinovirus demonstrated that antibodies need not induce conformational changes for neutralization and that viruses do not conceal receptor-binding regions from immune recognition. Ross River and Sindbis virus complexes were an early example of using antibodies to demark receptor-binding regions. The structure of an antibody bound to mouse norovirus is an example of antibodies binding to sharp protrusions on flexible receptor-binding domains. Finally, the structure of cucumber mosaic virus bound to a loop involved in aphid transmission demonstrated the importance of the context of antigen presentation and what happens when an antibody binds near an icosahedral symmetry axis.

Antibody neutralization of human rhinovirus

Rhinoviruses are the major causative agents of the common cold and cost the United States economy approximately $40 billion per year [1]. Therefore a vaccine to prevent or ameliorate the symptoms of the common cold is of great interest. Rhinoviruses are members of the picornavirus family that are characterized by non-enveloped capsid with a diameter of ~300 Å containing a single stranded, plus-sense RNA genome [2]. Other members of the picornavirus family include foot and mouth disease virus (FMDV), poliovirus, encephalomyocarditis virus (EMCV) and hepatitis A. The capsids exhibit pseudo T=3 icosahedral symmetry and are composed of 60 copies of the four capsid proteins: VP1, VP2, VP3, and VP4. VP1-VP3 have an eight-stranded anti-parallel beta-barrel motif structure and form the outer surface of the capsid while VP4 lies at the interface between the capsid and the interior genomic RNA [3]. VP4 is approximately 70 amino acids in length and is myristoylated at the N-terminus [4,5].

Antibodies are the major line of defense against picornavirus infections. In the case of HRV14, a number of studies have been performed to detail the antibody recognition and neutralization processes [6]. It had been long suggested that antibodies neutralize viral infectivity by inducing large conformational changes in the capsid. If this were the case, then the implication is that antibodies not only have to bind to the capsid but also induce large conformational changes to inactivate the virions. Further, it also suggested that viruses (such as FMDV) that have antigenic regions removed from the viral surfaces via flexible tethers could avoid antibody neutralization. To directly test this, the cryo-TEM structures of HRV14 complexed with the Fab fragments from three neutralizing antibodies (Fab17-IA, Fab12-IA, and Fab1-IA) were determined (Figure 1) [7,8]. Even though all three antibodies bind to the NIm-IA site (residues 91-95 of VP1) mAb17 and mAb12 are both strongly neutralizing antibodies while mAb1 is a weakly neutralizing antibody. It should be noted that Fab’s generated from representative mAb’s that bind to all four NIm sites (including Fab17) neutralized HRV14, albeit at higher ED₅₀’s [9] In all cases, none of these antibodies appeared to induce noticeable conformational changes in the capsid. However, it could be argued that, since Fab12 and Fab17 were likely to be able to bind bivalently to the viral surface, perhaps an intact IgG would distort the capsid when binding with both Fab arms. To that end, the cryo- TEM structure of mAb17 complexed with HRV14 was determined [10]. As with the Fab complexes, no notable conformational changes were observed. However, since all of these cryo-TEM structures were of limited resolution (~20Å), it was possible that smaller conformational changes went undetected. To directly test for this, the Fab17/HRV14 complex was crystallized and its structure was determined to ~4Å resolution (Figure 2) [11]. This structure clearly demonstrated antibodies do not need to induce conformational changes in the virions in order to neutralize infectivity. These results suggested that the major in vivo role of antibodies is bind to virion and work synergistically with other immune system components [12]. This crystal structure also demonstrated that antibody recognition is more plastic than previously thought in that it is able to bind into the relatively narrow receptor-binding region of the canyon [11]. Since the antibody makes direct contact with the receptor-binding region, this structure also demonstrates that viruses do not hide key receptor binding residues within folds of the virion surface. Indeed, most viruses do not need to hide from the immune system of a particular host since they do not establish persistent infections but rather just jump the next immunologically naïve victim.

Figure 1.

Composite picture of the structures of several antibody/virus complexes. In all cases, the antibodies are represented in various colors while the capsid surface itself is shown in grey.

Figure 2.

Details of two very different antibody/virus contacts. The ribbon diagram on the left represents the crystal structure of the Fab17/HRV14 complex. In this case, the hypervariable loops on the Fab cover the NIm-IA site and the heavy chain makes extensive contacts with the north and south walls of the receptorbinding region. On the right is hybrid structure of MNV-1 using the cryo-TEM structure of the MNV/Fab complex and the atomic structure of the MNV P domain. In this case, the epitope lies on a sharp protrusion to which the antibody makes contact.

While these results simplified the goal of creating a synthetic vaccine by focusing on a capsid recognition rather than possible antibody-induced conformational changes, developing synthetic vaccines against all 100 serotypes of HRV remains a daunting task. As shown in the structures of HRV14/antibody complexes, the antibodies make extensive contacts with the surface of the capsid that is not limited to a single antigenic loop [8,11]. Further evidence for this extensive contact is that antibodies to peptides corresponding to antigenic NIm loops fail to neutralize the virions [13,14], and antibodies raised against intact capsids do not bind effectively to peptides corresponding to NIm-IA loop (Smith, unpublished results). One notable exception is the case of HRV2 where there is cross-reactivity between the NIm-II site of the virion and a synthetic peptide [15]. Nevertheless, developing a repertoire of peptides representing the entire antigenic ensemble of HRV’s is not only impractical but also unlikely to elicit neutralizing antibodies.

Using a very different approach, the dynamic nature of human rhinovirus 14 (HRV14) was analyzed using limited proteolysis and mass spectrometry (MALDI) analyses [5]. In these experiments, the virus was treated with both matrix-bound and soluble forms of trypsin for varying periods of time and the resulting proteolytic fragments were identified by MALDI. Surprisingly, the Ntermini of VP4 and VP1 were found to be the most proteolytically sensitive portions of the capsid in spite of being buried inside the viral capsid. The antiviral ‘WIN’ compounds have been previously shown to stabilize the virions against thermal and acid denaturation. When added to the digestion mixture, these WIN compounds did not affect the intrinsic proteolytic activity of trypsin but protected the VP1 and VP4 termini from proteolysis for an extended period. Together, these results suggested that HRV14 is transiently exposing these termini in ‘breathing’ process and that the empty hydrophobic drug-binding region apparently plays an important role in these conformational changes.

As alternative approach to antibodies raised against intact capsids, work next focused on these VP4 and VP1 termini that are transiently exposed from the HRV capsid. The N-terminal region of VP4 is remarkably conserved among all of the HRV serotypes compared to the other capsid proteins, regardless of whether that portion of the capsid protein is exposed to immune surveillance [16]. Polyclonal antibodies against VP4 N-terminus were found to neutralize viral infectivity in-vitro [16]. In addition, the HRV14 VP4 antisera cross-reacts with other serotypes of rhinovirus: HRV16, and HRV29. Antibody neutralization closely paralleled MALDI analysis [17] in that antibody neutralization and proteolysis is enhanced at 37 ºC in the case of HRV16 whereas the elevated temperatures are not required for either phenomenon in the case of HRV14 and HRV29. Epitope mapping of the N-terminal 30 residues of VP4 suggested that the termini adopts a non-linear conformation. This was further substantiated by results showing that all of the copies of VP4 in a VP4 Ser5Cys HRV14 mutant at room temperature form cysteine cross-linked dimers. This cysteine cross-link does not form at 4ºC, suggesting that capsid breathing is essential for VP4 exposure and interactions. It was also shown that VP4 dimerization does not affect viral infectivity and therefore it seems likely that VP4 extrusion is a normal part of the cell attachment and entry process of rhinovirus. Together, these results suggest that VP4 might be useful as a pan-serotypic rhinovirus vaccine but it seems likely that better understanding of the VP4 oligomeric structure will be necessary for further optimization.

While these results seem encouraging that a peptide vaccine to the common cold may be possible, there are a number of issues that need to be resolved. Firstly, the HRV neutralization by the polyclonal and monoclonal anti- VP4 antibodies is not nearly as robust as that observed by mAb17-IA and mAb12-IA. However, weakly neutralizing antibodies such as mAb1-IA do have the same apparent efficacy as the polyclonal antibodies when applied to HRV16. Secondly, the time course of recognition/neutralization is much slower in the case of some serotypes (e.g. HRV16) than those that bind to external capsid sites. From the MALDI results, this is likely due to the rate of capsid breathing. It is unclear what impact this time-dependency will have on in vivo efficacy. Essentially, the antibodies need to ‘lie in wait’ for the N-termini to extrude from the capsid in order to bind and neutralize. Thirdly, it is not at all clear what structure of VP4 is being presented exterior to the capsid surface. Antibody recognition of VP4 peptides is markedly sensitive to length and, likely, conformation in solution. Further work is necessary to determine if there is an optimum length of VP4 peptide for inoculation and what conformation best mimics the VP4 N-terminus once exposed from the capsid. Finally, the problem with developing any vaccine to the common cold will be in vivo testing. Humans are the only mammals that exhibit symptoms upon HRV infection and it is not known what level of neutralization in-vitro is necessary to ameliorate symptoms in vivo.

Antibody binding to murine norovirus (MNV)

Murine norovirus (MNV-1) [18-20] is a member of the Norovirus genera of the Caliciviridae. Calicivirus particles contain 180 copies of the 56-76 kDa major capsid protein (Orf2) that is comprised of the internal/buried N-terminus (N), shell (S) and protruding (P) domains [21,22]. The S domain, an eight-stranded β-barrel, forms a ~300Å contiguous shell around the RNA genome. A flexible hinge connects the shell to a “protruding” (P) domain at the C-terminal half of the capsid protein that can be further divided into a stem region (P1) that connects the shell domain to a globular head region (P2).

The cryo-TEM structure of MNV-1 was initially determined to a resolution of ~12Å [23]. It was found that, compared to rNV virus-like particles (VLPs) [22] and San Miguel sea lion virus SMSV [24,25], the protruding domains are rotated by ~40º in a clockwise fashion and lifted off the shell domain by ~16Å. This work was subsequently followed by an ~8Å cryo-TEM structures of infectious MNV-1 and the molecular envelope was interpreted by the 2.0Å structure of the P domain [26,27]. At the higher resolution, a clear connection between the P1 domain and the shell domain in all three capsid subunits was observed. Also unlike the smooth protruding domains of rNV, MNV-1 has two clear ‘horns’, not dissimilar to those observed for the sapoviruses [24,25]. From fitting the MNV-1 P domain atomic structure into this envelope, the ‘horns’ are formed from loops A’-B’ and E’-F’ observed at the tips of the P domains. The connections between the P1 and S domains were of sufficient quality to build a basic backbone model by uncoiling the linker region.

With this hybrid model, the interactions between a neutralizing antibody and the P domains were examined (Figure 1, 2). Antibody contact was localized to a portion of the A’ β-strand and A’-B’ loop (aa 294 - 303) as well as the E’-F’ loop (aa 379 - 388) that form the ‘horns’ on the P2 domain. In this model, the ‘horns’ of the P2 domain fit snugly into the cleft between the heavy and light chain hypervariable regions and makes contact with all six antigen binding loops. In our previous analysis using the structure of the rNV P domain [23], there was an additional possible contact with parts of the C’-D’ loop and D’ β-strand (aa 345 - 358). However, the extension of the ‘horn’ pushes the antibody away from making contacts with the flatter regions of the P2 head. Therefore, the epitope for this antibody is clearly limited to the outermost tips of the P2 domain formed by the A’-B’ and E’-F’ protrusions. This is significantly different than a number of other antibodies that contact a broader, flatter surface (e.g. [11]). It is interesting that the antibody contact is limited to these loops and that these have two distinct conformations in the crystal structures [26]. While it is not unexpected that the antibody contact area includes the known escape mutation site to this antibody (L386) [28], it is interesting that it also overlaps the attenuation site, E296 [19]. This is akin to studies on other viruses demonstrating that epitopes and regions of viral capsids that change during host adaptation can overlap (e.g. [29,30]).

It is absolutely clear that the hinge region between the S and P domains affords a remarkable degree of flexibility in the P domains. The simplest explanation for the role of this transition is that it gives the P domains flexibility that may be used to optimize interactions with cell receptors during attachment and entry. In this way, the P domains can increase their avidity for the cell surface by being more facile in adapting to the presentation of cellular recognition motifs. With regard to antibody neutralization, this location of antibody binding is very far removed from the core structural elements of the virion. Not only are the P domains highly motile and far removed from the shell, but the antibodies themselves are binding to an outermost extremity of the P domains. Therefore, as with most antibody virus interactions, it seems highly unlikely that antibodies can or need to induced conformational changes in the capsid for neutralization.

Antibody neutralization of the alphaviruses

The alphaviruses are a group of 26 icosahedral, positive-sense RNA viruses that are primarily transmitted by mosquitoes [31]. These ~700Å diameter viruses are some of the simplest of the membrane-enveloped viruses, and members of this group cause serious tropical diseases [32]. The viral RNA genome and 240 copies of the capsid protein form the nucleocapsid core [33-38], and the E1 and E2 glycoproteins form heterodimers that associate as 80 trimeric spikes on the viral surface. E1 has a putative fusion domain that may facilitate host membrane penetration [39,40]. E2 contains most of the neutralizing epitopes and is also probably involved in host cell recognition [41-43].

To examine the mechanism of antibody-neutralization and to identify the portion of E2 involved in receptor recognition, two antibodies were examined: SV209 and T10C9 (Figure 1) [44]. Antiidiotyic antibodies to SV209 compete with Sindbis virus (SINV) for its cellular receptor and block viral attachment by ~50% [42]. This implies that the original SV209 antibody is recognizing at least a portion of the spike involved in cellular recognition. The naturally occurring mutation in Ross River virus (RRV) that facilitates escape from the T10C9 antibody maps to residue T216 of E2 [45]. This residue is presumably near the cell receptor-binding site since residue N218 was found to vary as the virus adapted to growth in chicken cells [46] and residue T219 mutates to ALA during the course of an epidemic in humans [47].

In both virus/Fab reconstructions (Figure 1), the Fab fragments were observed to bind to the outermost tips of the timeric spikes [44]. When compared to reconstructions of the virus alone, the binding of the antibody did not appear to cause conformational changes in the virion. While the two antibodies bound to their respective viruses with markedly different orientations, their binding footprints were nearly identical on the highly exposed tip of the spike. Further, the results with the antiidiotypic antibodies suggest that there is a direct overlap between the antibody and receptor contact areas.

Antibody recognition of cucumber mosaic virus (CMV)

Cucumber mosaic virus (CMV), the type member of the genus Cucumovirus (family Bromoviridae), infects over 800 plant species and causes economically important diseases of many crops worldwide [48]. CMV is transmitted by aphids in a non-persistent manner; it does not circulate or replicate in the aphid.

The X-ray crystal structure of CMV revealed an exposed βH-βI loop [49] the sequence of which is highly conserved among strains of CMV and other cucumoviruses [50]. Mutations in several of the loop residues (D191, D192, L194, and E195) had no significant affect on virion formation or stability but they did reduce or eliminate aphid transmission [50]. To better understand the molecular basis for virus transmission by insects, antibodies were developed against this loop and the structure of this virus-antibody complex was determined [51].

The cryo-TEM and modeling results clearly demonstrated that this antibody binds immediately adjacent to an axis of icosahedral symmetry and only one Fab binds per penton (Figure 1). Indeed, each antibody bound to several antigenic loops at the same time. While antibodies are known to be able to bind simultaneously to multiple viral subunits (e.g. [52,53]), this was the first example in which an antibody bridges the same regions of two or more identical and adjacent capsid subunits. Since the βH-βI loop that is thought to interact with the receptor molecule in the aphid, it may be that the aphid receptor, similar to antibody 3A8-5C10, also exhibits quasi-equivalent specificity and may interact only with pentons or only with hexons. This is a clear demonstration that the context of an antigenic site is just as important as the structure of the antigenic site itself.

Conclusions

These structural studies have shown several important aspects of antibody recognition and neutralization. Antibodies are remarkably adept at binding to a wide range of surfaces; they can bind to sharp protrusions on flexible domains, dig down deep into canyons and crevasses, or bind bluntly to a relatively flat surface. They do not appear to cause conformational changes in the virion upon binding and neutralization. Indeed, the energy required to cause such conformational changes would necessarily affect binding affinity. More importantly, if conformational changes are required for neutralization, then viruses would have most certainly developed floppy antigenic loops to isolate the effects of antibody binding. However, as is the case with the antibodies to the transiently exposed N-termini of HRV14 VP4, it is possible that antibodies can ‘lie in wait’ for a particular conformation to be presented. The problem with developing vaccines of this nature will be in determining what structure to present and even then the efficacy will be impacted by frequency that this alternative conformation is displayed.

Bullet Points.

1. Antibodies do not cause conformational changes in viral capsids upon neutralization.

2. Receptor binding regions are not hidden from antibody recognition.

3. It may be possible to target vaccines to conserved portions of viruses that are dynamically exposed during capsid breathing.