Biophysical data from AFM, cryoEM, dynamic light scattering and sedimentation experiments show a significant improvement in the morphology, homogeneity and thermal stability of the HPV16 VLPs upon D/R treatment [12, 37]. Antigenicity, which can be measured in vitro, is an accurate and convenient metric for product quality and stability, as a surrogate marker for in vivo immunogenicity of the VLP. For quantitative antigenicity analysis, sandwich ELISA has high specificity and sensitivity and can be developed for VLP characterization and product release with the desired mAbs. However, for bioprocessing or stability studies of final vaccine products, the interpretation of sandwich ELISA data can be complicated as both the capture Ab and detection Ab contribute to the final assay signals. Therefore, to understand the impact of D/R on VLPs for different epitopes, a set of fluorescence-based competitive ELISA assays with high specificity and sensitivity, allowing VLPs to freely interact with a given mAb in solution, was developed for probing epitope specific antigenicity on VLPs [28, 29]. The quantitative ELISAs with half-maximal inhibitory concentration (IC50), by putting mAbs one at a time in the assay, yielded quantitative information on the impact of D/R treatment on individual epitopes.
Among an array of immunochemical assays for HPV VLP epitope characterizations, including competition ELISAs (IC50), sandwich ELISA, equilibrium dissociation constant determination, relative antigenicity and pair-wise epitope mapping, the epitope specific relative IC50, or rIC50, assay is the most sensitive assay with straightforward data interpretation due to the use of a single mAb in the assay. A ~3-fold enhancement in IC50 was observed for H16.V5 after D/R treatment of VLPs. In the relative antigenicity assay by SPR, the enhancement was ~2-fold. In addition, in a sandwich ELISA in which VLPs were captured with H16.J4 and detected with H16.V5, a moderate 30%-50% increase was observed [19, 20]. This is not surprising given that these two mAbs showed opposing effects for binding to the post D/R VLPs. The sandwich assay implies a more dominant contribution from the detection H16.V5 since an overall increase of antigen content was seen, and conversely a smaller contribution by H16.J4 due to poorer capture of post-D/R VLPs (Figure 3B). Efficient D/R was shown to yield more fully assembled and presumably more virion-like VLPs with greatly reduced binding to H16.J4 and H16.O7, while rendering H16.H5 binding completely undetectable. This is consistent with a study by Culp et al.  in which pseudovirion binding was essentially undetectable for H16.H5, H16.J4, H16.O7 and H16.I23 while binding of each of these antibodies to L1-only VLPs derived from insect cells was readily detectable. It is thus conceivable that the epitopes for H16.H5 and H16.I23 were initially exposed in the imperfectly assembled VLPs after purification (from yeast or insect cells), and subsequently became inaccessible or buried after the formation of virion-like seamless VLPs after D/R during bioprocess or in the L1 + L2 pseudovirions. Since the H16.I23 epitope is not close to any intercapsomeric interfaces, the loss of H16.I23 binding to post-D/R VLPs and pseudovirions may be due to the rigidification of loops on the capsomere surface that may accompany particle assembly thereby restricting accessibility to the epitope. Our observations also caution against interpreting data from the widely used sandwich ELISA, particularly with respect to the dis- and reassembly of VLPs, where opposing and convoluted effects could be seen on the capture Ab and detection Ab.
The events that are likely to occur during D/R treatment on purified VLPs are: (1) breakdown of non-specific aggregates including disulfide-bonded aggregates (an overall reduction in particle size and increase in monodispersity were observed by dynamic light scattering and analytical ultracentrifugation); (2) unmasking of some virion-like epitopes previously hidden due to non-covalent or covalent association with other VLPs or capsomeres through aggregation or non-native disulfide bond formation; (3) promotion of complete assembly of the closed icosahedral VLPs; (4) reduction of non-native disulfide bonds and formation of native and thermodynamically preferred intra- and intercapsomeric disulfide bonds; (5) formation of native electrostatic and hydrophobic interactions upon correct capsomere assembly by formation of intercapsomeric disulfides [40, 41]. Making VLPs with epitopes resembling those of authentic virions is essential in vaccine production. Understanding and quantitating the epitopes with multiple non-overlapping mAbs on the recombinant VLPs may provide insights on the understanding of not only serotype specific protection for vaccine types, but cross protection of non-vaccine types, by current vaccines .
Mapping of the previously identified antibody epitopes onto the HPV16 VLP atomic model shows that intercapsomere contacts in the VLP can explain the observed changes in antibody reactivity upon D/R and allow more precise and more consistent mapping of the antibody epitopes. Within the H16.H5 epitope (residues 174-185), residues 175, 177 and 178 are occluded by the C-terminal arm of an adjacent capsomere in the HPV16 VLP but not in pentameric capsomeres, suggesting that the observed reduction of H16.H5 and H16.O7 antibody reactivity to VLPs upon D/R is due to the occlusion of one or more of these residues. Cys175 forms an intercapsomeric disulfide with Cys428 in the C-terminal arm of a neighboring capsomere [39, 43]. When the VLPs are fully assembled during D/R treatment, solvent access to the EF loop bearing Cys175 is thus occluded. VLPs that bind to H16.H5 antibody are therefore likely to have incomplete intercapsomeric disulfide formation, to be missing some capsomeres, or to be distorted such that H16.H5 can still access the epitope. However, residues 175 and 177 are widely conserved across HPV strains. Since the observed reduction of H16.H5 and H16.O7 binding to VLPs upon D/R is limited to HPV16, the reduction in binding cannot be fully explained by the occlusion of residues 175 and 177. Residue 178, which is a valine only in HPV16, is the only unconserved residue to be occluded upon VLP assembly. We therefore propose that Val178 plays a key role in H16.H5 and H16.O7 binding to misassembled (pre-D/R) VLPs or to unassembled HPV16 L1 in GST-L1 fusion protein  or in individual capsomeres, and that the occlusion of Val178 upon correct VLP assembly is largely responsible for the loss of reactivity against these antibodies. Moreover, we note that in our atomic model Ser173 forms a hydrogen bond with His431 from the C-terminal arm of an adjacent capsomere (Figure 8A, B). This interaction results in the occlusion of Ser173 upon VLP assembly. Since the Ser173-His431 pair is present only in HPV16, Ser173 may also be part of the H16.H5 and H16.O7 epitopes, even though Ser173 lies just outside the previously identified linear epitope of H16.H5 and H16.O7. The occlusion of Ser173 upon VLP assembly would provide an additional explanation for the reduction in antibody reactivity. We note that epitopes that are near interacapsomeric interfaces, such as those of H16.H5 and H16.O7, are more likely to exhibit antigenic differences between the 40-nm and 55-nm particles since structural differences between the two types of particles are likely to be concentrated at the interfaces between capsomeres. However, there is no evidence of any antigenic differences between the two types of particles.
In contrast to H16.H5, VLP binding to antibodies H16.V5, H16.E70 and H263.A2 is enhanced upon D/R treatment. The 3.6-Å resolution cryoEM structure of BPV1 and our atomic model of HPV16 do not, however, suggest that any epitopes become exposed upon virus assembly . We conclude that the enhanced binding to post-D/R VLPs by H16.V5, H16.E70 and H263.A2 is most likely due to the decrease in aggregation in post-D/R samples, which was observed by dynamic light scattering with much smaller polydispersity indices (unpublished data) and by AFM and EM images [12, 16]. We note that the epitope of these three antibodies contains an unpaired cysteine, Cys345, which is located within a few Ångström of the outer surface of the capsomere. Cys345 may be exposed to solvent in misfolded or improperly assembled VLPs, which could cause VLP aggregation through the formation of non-native disulfides.