Binding and neutralizing activities of an antibody panel
Human IgG1 mAbs were obtained by screening, sequencing, and recombination of single B cells isolated from vaccinated individuals. A panel of antibodies against four types of antigens (15 HPV6-specific mAbs, 11 HPV11-specific mAbs, 11 HPV16-specific mAbs, seven HPV18-specific mAbs, and eight HPV6/11/16/18-cross-binding mAbs) was established. Enzyme-linked immunosorbent assays (ELISAs) were used to detect the binding activities of HPV mAbs to HPV VLP antigens. After antibody dilution, optical density (OD) values at 450/620 nm for each concentration were entered into Prism 8 software. The four-parameter fitting model was used to calculate the half-maximal effective concentration (EC50) value for each mAb; a smaller EC50 value revealed that the mAb had stronger antigen-binding ability. As shown in Supplementary Fig. 1, among HPV-specific mAbs, the EC50 values for F5-77, F5-187, F5-196, and F5-203 were all below 20 ng/ml, indicating that these mAbs had strong antigen-binding ability and could be used to identify mAbs suitable for assessment by sandwich ELISA. Because the immunoprotective effects of preventive HPV vaccines depend on the production of neutralizing antibodies, measurements of HPV VLP binding and neutralizing activities in candidate mAbs have become important aspects of vaccine efficacy assessment. The pseudovirion-based neutralization assay (PBNA) is regarded as the gold standard for clinical detection of neutralizing activity in mAbs. In this study, the neutralizing activities of mAbs were detected by PBNAs, which provided reference data for mAb screening via sandwich ELISA. A smaller half-maximal inhibitory concentration (IC50) value revealed that the mAb had greater neutralizing activity. As shown in Supplementary Fig. 1, among HPV6-specific mAbs, the IC50 values of F5-77, F5-187, F5-196, and F5-203 were all below 10 ng/ml, indicating that these mAbs had robust neutralizing activity.
Conformational sensitivity of mAbs
The integrity of key epitopes on HPV VLPs is essential for vaccine efficacy. To ensure that the resulting mAbs reflected conformational changes in HPV VLP antigens, thereby improving quality control, this study analyzed the conformational sensitivities of mAbs to HPV VLPs. HPV VLP antigens were incubated overnight at 37°C with dithiothreitol (DTT) to denature their structures, then subjected to indirect ELISA with various mAbs12. Each antibody was diluted to 2 µg/ml, and the various reactivities of mAbs against denatured and intact VLPs were compared using OD450 values. An antibody was considered conformationally sensitive if it exhibited an OD value greater than 1.0 for intact VLPs and less than 0.1 for denatured VLPs. F5-77, F5-187, F5-196, and F5-203 were identified as conformationally sensitive antibodies (Fig. 1a, b).
Immunodominance of mAbs in human anti-HPV serum
When evaluating vaccine antigen efficacy in vitro, the epitope recognized by an antibody should be identical (or similar) to the epitope recognized by antibodies produced by the human body after vaccination (i.e., the dominant epitope after immunization). Because neutralizing antibody titers are low in the serum of individuals naturally infected with HPV, the present study used an HPV detection antibody to block the HPV antigen binding of clinical serum samples (ethical approval number: 2019-043) from 57 patients immunized with 9-valent HPV vaccine; the resultant blocking rates were used to compare dominant antibody proportions in human serum. The positivity threshold was set to a blocking rate of 30%, as previously reported11. The blocking positivity rates of F5-77, F5-187, F5-196, and F5-203 were 100%, 89%, 100%, and 93%, respectively, indicating that they recognized dominant epitopes on the surface of HPV L1 VLPs (Fig. 1c–f).
Antibody pairing of double antibody sandwich ELISA
HPV6, 11, 16, and 18 cross-binding antibodies were selected as capture antibodies; HPV6, 11, 16, and 18 type-specific antibodies with high binding activity, strong neutralizing activity, conformational sensitivity, and a high blocking rate in human serum were selected as detection antibodies for horseradish peroxidase (HRP) labeling. These antibodies were paired as follows. First, HPV VLPs were diluted to 200 ng/ml and an HRP-labeled antibody was diluted to 1 µg/ml; next, both components were mixed with chromogenic solution and the OD value was recorded. Paired mAbs with a signal-to-noise ratio of > 20 were selected for subsequent screening. As shown in Fig. 1g, F5-127 exhibited good binding to F5-77, F5-187, F5-196, and F5-203; therefore, it was selected as a capture mAb.
Versatility of mAbs
Coated antibodies paired with enzyme-labeled antibodies were used to detect HPV VLP antigens (including VLPs produced by yeast, E. coli, and insect cells) from seven HPV vaccine manufacturers, and EC50 values were calculated. Detection mAb versatilities were defined as the ratio of the maximum EC50 value to the minimum EC50 value for each manufacturer; paired mAbs with a ratio of < 4 (Fig. 1h–k and Supplementary Fig. 2) were selected as candidate paired mAbs for the kit. Based on the previous screening results, the type-specific mAbs F5-77, F5-187, F5-196, and F5-203 (all exhibiting high binding activity, strong neutralizing activity, conformational sensitivity, immunodominance, and detection versatility) were selected as detection mAbs for HPV6, 11, 16, and 18, respectively.
Analytical performance of IVRP method
To verify the accuracy of the new method for determining the relative potency of HPV VLPs in vitro, we selected five initial pre-dilution concentrations of HPV VLP (i.e., 50%, 75%, 100%, 150%, and 200%), and determined their activity. Each titer was measured twice by four experimenters at different times. According to the EC50 values of the reference and various pre-dilution concentrations, the relative efficacies of the actual measurements of different concentrations of HPV VLPs were calculated, and the results of eight experiments were reported (Supplementary Table 1). As shown in Fig. 2, the relative bias of each titer relative to the measured titer value was in the range of 30%. The geometric coefficient of variation (%) of each titer relative to the measured titer value was less than 30%. The log(abscissa) of the theoretical value of the titer was used to plot the log(ordinate) of the measured value of the titer, and the least squares method was used to perform linear regression. The correlation coefficients of the linear regression equations all exceeded 0.98. Specificity refers to the ability of a measurement method to distinguish related substances. This metric was analyzed during the addition of different types of HPV antigens to a single type of test antigen (i.e., to assess the abilities of mixes containing four types of HPV antigens (6/11/16/18) and type-specific detection mAbs to detect the titers of the other three antigens). The four antigens were mixed at concentrations of 100, 400, and 1000 ng/ml with a ratio of 1:1:1:1. Relative bias was measured, and specificity was evaluated (values < 30% were preferred). The experiment was repeated three times.
Structures of HPV L1 pentamer and antibody complexes
To elucidate the epitopes, binding characteristics, and neutralization mechanisms of the four selected antibodies, we conducted structural analysis of the HPV L1 and antibody complexes. The high heterogeneity of HPV VLPs hinders high-resolution structural determination. Therefore, we depolymerized HPV VLPs into homogeneous and stable HPV L1 pentamers using the deoxidizing agent DTT. HPV L1 pentamer and antibody complexes were prepared by mixing HPV L1 with individual Fab fragments at a molar ratio of 1:1.2 for cryo-electron microscopy (cryo-EM) investigation. The cryo-EM structures of the F5-77, F5-187, F5-196, and F5-203 complexes were determined at resolutions of 4.3, 3.1, 2.9, and 2.9 Å, respectively (Fig. 3 and Supplementary Fig. 3). The high-quality electron density maps generated by cryo-EM allowed us to build models of these four complexes (Supplementary Fig. 4). We observed two distinct binding patterns for the four antibodies: five F5-196 Fabs were clustered around one pentamer, whereas one Fab molecule was vertically bound to the central region of one L1 pentamer for the other three antibodies (Fig. 3a–d, left panels). Similar binding modes were previously reported13,14. In the broader context of the capsid structure, these four antibodies targeted conformational epitopes within the pentamer (Fig. 3a–d, right panels, and Supplementary Fig. 5), rather than epitopes along the edges of the pentameric blocks of the capsid. Thus, they recognized any forms of VLPs comprising various numbers of L1 pentamers from different expression systems, offering a structural explanation for the universality of these IVRP candidate antibodies. Specifically, F5-77, F5-187, and F5-203 bound to one side of the central cavity of the pentameric plateau, where they contacted 2–3 copies of the neighboring L1 (Fig. 3a–d, right panels, and Supplementary Fig. 4). Previous studies have indicated that the minor capsid protein L2 resides within the axial cavity of the L1 pentamer and mediates HPV infection through exposure of its N terminus, which is cleaved by the proprotein convertase furin15–17. Therefore, these three antibodies might inhibit HPV infection by blocking conformational changes on the L1 capsid necessary for infection or by blocking L2 N-termini that protrude onto the capsid surface, subsequently preventing recognition and cleavage by furin. However, five F5-196 Fabs were clustered on the top and rim of one pentamer; one Fab interacted with a copy of L1. In total, 360 F5-196 Fabs were bound to VLPs, occupying nearly the entire surface of the HPV16 capsid; this binding may prevent HPV16 from attaching to cellular receptors. Additionally, the saturated occupancy of F5-196 is consistent with the experimental observation of up to 90% inhibition against vaccinated human sera.
The tight binding between HPV6 L1 and F5-77 was achieved through a network of hydrophobic interactions formed by F112, G123, S124, P128, and G129 from HPV6 L1; Y33, N54, R100, Y101, and Y104 from the heavy chain; and hydrophilic interactions, including 13 hydrogen bonds. The epitope for F5-187 contained 10 residues, which were primarily located in the L1 DE-loop (Y123, D124, N128, G130, Y132, G134, and N135), FG-loop (N278 and N279), and HI-loop (S354) of HPV11. Extensive hydrophilic interactions contributed to the high binding affinity between F5-187 and HPV11. Similar to F5-77, interactions between HPV18 L1 and F5-203 consisted of a network of hydrophobic interactions formed by P121, F122, A134, V139, and M282 from HPV6 L1; F27, Y35, Y50, W92, and W99 from the light chain; Y104 and I106 from the heavy chain; and hydrophilic interactions, including 11 hydrogen bonds. However, the epitope of F5-196 mainly included residues N138 and A139 of the HPV16 DE-loop; residues N270, K278, S280, G281, and N285 of the HPV16 FG-loop; and residue K356 of the HPV16 HI-loop (Fig. 4a and Supplementary Table 2). The epitopes of these antibodies were mainly distributed on the HPV L1 DE, FG, and HI-loops, which are also often mapped as neutralizing epitopes13,14. Based on the amino acid sequence conservation analysis of 9-valent HPV types, the type-specific mAbs F5-77, F5-187, F5-196, and F5-203 primarily bound to the most variable regions of the HPV L1 DE, FG, and HI-loops (Fig. 4b and Supplementary Fig. 6). The sequence specificities of binding sites targeted by these four antibodies determine their type-specific recognition and neutralization. Furthermore, in a comparative analysis of the structural conformation of the main binding regions of the DE, FG, and HI-loops among the four specific antibodies targeting HPV L1, F5-77, F5-196, and F5-203 predominantly bound to positions within the DE/FG-loop, which also displayed significant conformational differences. Similarly, F5-187 primarily bound to a position with substantial conformational variation in the FG-loop (Fig. 4c, dashed box). Collectively, these antibodies exhibited high conformational sensitivity because they mainly interacted with regions exhibiting structural plasticity on HPV L1.
Correlation between in vitro and in vivo potency
The quadrivalent HPV vaccine, Gardasil®, was incubated at 56°C for 0, 1, 6, and 12 h (designated groups 1 to 4, respectively), and subjected to different degrees of heat-accelerated destruction. Then, the in vivo efficacy and in vitro relative efficacy were measured simultaneously. As shown in Fig. 3, after immunization with vaccines subjected to different degrees of destruction, the neutralizing antibody titers in mouse sera showed a decreasing trend for vaccines against HPV6, 16, and 18; there was no significant difference for the vaccine against HPV11. The in vitro relative efficacy exhibited a decreasing trend for vaccines against HPV6, 11, 16, and 18 after different degrees of destruction. Correlation analysis was performed between the in vivo efficacy and in vitro relative efficacy. As shown in Fig. 5, the correlation coefficients (R2 values) between in vivo and in vitro efficacies for vaccines against HPV6, 16, and 18 were 0.89, 0.70, and 0.96, respectively; these values were greater than 0.5, indicating a good correlation. The correlation coefficient between in vivo and in vitro efficacies for the vaccine against HPV11 was 0.26, indicating a poor correlation. Nevertheless, this experiment provided strong evidence that the in vivo efficacy method established in this study can serve as a sensitive measure of vaccine quality. It also strongly supports the use of in vitro experiments as an alternative method that reduces the number of animals used while meeting the requirements of the 3R principle and ICH.
Collaborative validation
The four-type IVRP kits were distributed to six vaccine manufacturers for collaborative validation. The relative efficacies of three batches of HPV stock solution and three batches of finished products were simultaneously measured using the kit and each manufacturer’s in-house method; this experiment was repeated three times. As shown in Fig. 6, the in vitro relative efficacy test results of three batches of monovalent antigens and three batches of finished vaccines from six HPV manufacturers were all between 0.5 and 2.0; they were primarily concentrated around 1.0, approximating a normal distribution and indicating that the kit can be used to detect both vaccine stock solution and finished products. Comparison of the in vitro relative potency test results between the kit method and the manufacturers’ methods revealed that the biases of Lab1, Lab2, Lab3, Lab4, Lab5, and Lab6 in the three batches of stock solution from each manufacturer were acceptable at < 50%;. These findings revealed that the test results for the three batches with relative bias > 50% considerably varied