The analysis of antigenic differences is critical for FMD surveillance and vaccine strain selection. Emergence of antigenically different strains in FMD outbreak needs to be determined in relation to the respective vaccine strains using appropriate techniques . As neutralizing antibody titers correlate well with protection in animals, the virus neutralization test has been adopted since 1977  and is widely used as the reference test system for vaccine strain selection. Although the 2D-VNT has been considered the gold standard, significant variation has been described with the VNT results [6, 22]. Mattion  indicated that r1 value estimations using low serum titer become less precise. Therefore depending on only a single test for vaccine matching may not be suitable.
To better antigenically monitor new outbreak strains, a panel of FMDV/O specific mAbs was produced. Seven mAbs that reacted with the neutralizing sites and two mAbs that reacted with the non-neutralizing sites were included in this panel. Four mAbs reacted with recombinant proteins indicating binding sites of these mAbs were linear. The binding sites for mAb F12VP1O-2 and F21-64 identified using peptide ELISA are located on VP1 antigenic site 1b (aa198-213) and 1a (aa136-151). Although the mAb F12VP1O-2 binding site is located in the antigenic site 1b, it was unable to neutralize the virus in a VNT. A possible explanation may be that site 1a (the βG ± βH loop) and 1b (the carboxy terminus) on VP1 together involve antigenicity and receptor binding . The mAb F12VP1O-2 against only C-terminal would not be able to completely block virus attachment, thus demonstrating negative neutralizing activity.
In general, linear antibody binding epitopes can be located using overlapped peptides. Surprisingly, F11VP2O-2 and F21-48 failed to react with any peptides, even though they reacted with recombinant O/VP 1 and VP2 in an indirect ELISA. A similar observation has been reported previously . An explanation is that mAb reactivity with a continuous epitope not only depends on the amino acid sequence, but also on a stretch of contiguous residues to assume the correct conformation . The binding sites of the other neutralizing mAbs with conformational epitopes were identified using the mAb resistant mutant selection method. Though the mAb F21-48 reacted with a linear epitope, but it was unable to react with any VP1 peptides. The exact binding site was also identified using the mutant selection method. After the selection in the presence of high concentrations of mAbs, the neutralizing mAb resistant mutants were generated and genetically analysed. The nucleotide mutations causing one significant amino acid substitution in protein structures were determined. The substituted amino acid residue was considered the antigenic site recognized by the mAbs. Combining the results obtained from the peptide ELISA and the mutant selection method, we concluded that the binding sites of seven mAbs are located on all five previously identified antigenic sites on FMDV/O involving VP1, VP2, and VP3 major capsid proteins [11, 19, 26, 27]. It was confirmed that these sites were exposed in the 3D model of protein structure.
Using this well characterized mAb panel, an antigenic profiling ELISA was developed with a modification from previously reported ELISA . To correct the amount of virus trapped by the capture antibody, a serotype independent mAb (F21-42)  was used to standardize the ELISA results instead of a polyclonal serum. This mAb demonstrated a consistent reactivity to all virus isolates, while the polyclonal anti-FMDV/O mouse serum pool demonstrated poor binding to some isolates.
The forty-six FMDV/O field isolates were analysed using the mAb antigenic ELISA. The isolates ECU/4/10 and HKN/2/11demonstrated highest antigenic variation compared with other isolates. Four of the nine mAbs failed to react with these two isolates. In addition, one mAb demonstrated low reactivity to HKN2/11 and two mAbs demonstrated low reactivity to ECU4/10. It is assumed that various amino acid alterations occurred in/near the antibody binding sites. The mutated binding sites identified by the mAbs for ECU/4/10 are located on the antigenic sites 1a, 1b, 2, 4 and one on non-neutralizing site. This result of high antigenic variation is consistent with previously reported data for the ECU/4/10 isolates . Viruses circulating in Ecuador during the years 2009–2010 were examined using monoclonal antibody profiling. The results showed that the viruses lost reactivity with the four mAbs, three of them with neutralizing properties. Moreover, results obtained with in vivo challenge indicated a lack of protection by the vaccine virus (O1/Campos) .
Both the GH loop and the C-terminus of VP1 formed antigenic site1 (a, b) are highly exposed regions on virus surface . It has been indicated that the region of VP1200-273 is important to both antigenicity and receptor binding of FMDV [19, 29–31]. It is involved in cell attachment because selective removal of these residues resulted in FMDV particles no longer capable of binding to cells . In this mAb panel, a mAb F12VP1O-2 is able to recognize this region located on the C-terminal of VP1198-213. Five out of nine isolates demonstrated weak to no reactivity to this mAb indicating high variations at this mAb binding site. Unique amino acid substitutions for TAW/12/98 (aa 212-A) and KUW/1/11 (aa 205-M) were confirmed by sequence analysis. Those substituted amino acids are adjacent to the antigenic site 1b (VP1208) and might explain a low r1 value in 2D-VNT. Likely, this mAb can be used to monitor whether any mutation occurred in this location.
Correlation coefficients were used to determine the relationship between the 2D-VNT and the antigenic profiling ELISA. However, in current study, a poor relationship for 46 isolates between the two tests was observed (r = 0.085, p = 0.57). In this study, the 2D-VNT was carried out using a polyclonal serum obtained from the O1/Manisa vaccinated cattle. Whereas the panel of mAbs used in the antigenic ELISA was raised against the O1/BSF (O1/Campos) strain. A previous report indicated that O1/Manisa and O1/Campos strains are related but they were not a perfect match by the 2D-VNT (r1 = 0.64/0.62) . The sequence analysis revealed that the five antigenic sites identified by the mAb panel in O1/BSF are identical to O1/Manisa in antigenic sites 2, 3 and 4, but some amino acids changes were observed in site 1. To obtain better results for vaccine matching, mAbs raised against each specific vaccine strain should be used. In practice, this is impossible because of the large amount of work required for mAb production and characterization. Alternatively, a pooled of polyclonal sera from O1/BFS vaccinated animals should be used in the 2D-VNT to ensure the results are comparable; however, we were unable to obtain the pooled O1/BFS vaccinated polyclonal sera. In spite of this, it is unlikely that the poor correlation between the 2D-VNT and the antigenic ELISA is due to the fact that different strains were used to generate the detecting antibodies in each assay for two reasons. Firstly, despite the fact that the mAbs were raised against O1/BFS antigens, the locations of antigenic sites are common for all serotype O isolates. Thus any mutations that occur in these sites should be detectable using the mAb panel. Secondly, the r values of antigenic profiling ELISA for each isolate were calculated versus the reference strain (O1/Manisa). Similar finding was reported by Mahapatra et al.,  who used the antibodies against the same strain for mAb antigenic ELISA and VNT; they also observed a negative correlation between the two tests.
A possible explanation for the poor relationship between the 2D-VNT and the antigenic ELISA may be that mAbs are a monovalent antibodies and that a single mutation occurring in the binding site may cause poor mAb binding, but may not affect 2D-VNT results by polyclonal antibodies. Another possibility could be that in addition to the known sites, other undefined sites (neutralizing or non-neutralizing) may also be important in the induction of a protective immune response. This was supported by the fact that animals with low levels of neutralizing antibodies could also be protected [34–37]. A broad repertoire of epitope specificities following vaccination has been observed in previous studies [38, 39]. Also, Nagendrakumar  have reported that the hyperimmune sera collected from vaccinated animals were unable to compete with the panel of mAbs used in their study.
Current studies have shown that the antigenic profiling ELISA is not a substitute for the 2D-VNT in vaccine matching because of the poor correlation between the two tests. However, it may still be used as one of the parameters to measure how field isolates are antigenically related to a vaccine strain and can provide valuable information on why a vaccine strain and a field isolate do not match. A good example is that this approach has been used successfully to define the epitope mutations of ECU/4/10, although approximately a 10% sequence difference in VP1 with the vaccine strain O1/Campos has been reported . Our finding explained previous finding that why there are no matching vaccine strains for this isolate.
To improve the current approach, more mAb representative of each antigenic site and multi panels of well-defined mAbs against different strains should be included and used in antigenic ELISAs. More non-neutralizing mAbs should also be included in the panel since non-neutralizing antibodies might also contribute to protection [40, 41]. These will allow the antigenic ELISA results to be more compatible with the 2D-VNT.
A powerful screening method for characterising FMD viruses is the genetic sequencing of the capsid protein. The genetic sequence can reveal either emergence of new strains or how similar outbreak isolates are to vaccine strains. In order to see whether any association exists between amino acid variation and pattern of r1 value in 2D-VNT, deduced amino acids in the P1 region of 9 isolates with r1 < 0.3 were compared with a vaccine strain O1/Manisa. The isolate ECU/4/10 and ETH/39/09 display multiple unique amino acid substitutions and amino acid depletion on several neutralizing sites. The high sequence variation of these two isolates explained the low r1 value for 2D-VNT. All other isolates did not show significant amino acid alterations compared with the vaccine strain O1/Manisa on the antigenic sites. However, amino acid substitutions near antigenic sites were observed. The changed amino acids near antigenic sites may alter the structure of outer capsid proteins, rendering antibodies unable to bind to viruses and preventing reference serum from neutralizing the viruses.
It has been emphasized that the neutralizing site 1 (VP1140-160) plays a role in the antigenicity of FMDV. However the residue at position 139 located in the immunodominant region of VP1 (βG-βH loop) is also highly variable . Amino acid modification at position 139 contributes to serum neutralization resistance in a serotype O isolate . In the current study, four isolates, ETH/39/09, UAE 9/09, TAW/12/98 and VIT/32/11 showed amino acid substitution or deletion at position 139 of VP1 which may explain low r1 values in 2D-VNT. Similarly, Das et al.  noted that three isolates had an amino acid deletion at position 139 of VP1. This finding confirmed that this position (VP1139) is antigenically important. In future sequence analysis, attention should be paid to this region. Substitutions outside of the antigenic sites have been shown to play an important role in the antigenic diversification of FMDV [20, 44]. Although one strain (IRN 11/2006) demonstrated a low r1 (0.21) value in 2D-VNT, no unique amino acid changes were identified in/near any neutralizing site of the outer capsid proteins. A possible reason for the low r1 value in 2D-VNT might be that the polyclonal serum used in the 2D-VNT is not a serum pool. It has been suggested that a suitable reference serum for vaccine matching r1-value experiments should be a pool or a medium to high titer serum . The r1 value was higher than 0.3 when a different O1/Manisa vaccinated serum (National Centre for Foreign Animal Disease, Canadian Food Inspection Agency) was used in the 2D-VNT.
In comparison to the results of antigenic mAb ELISA and sequence analysis, multiple mAb binding reduction (sites 1, 2, 4) in ELISA and several amino acid alterations in the antigenic sites (sites 1, 3, 4) by sequence analysis were observed for isolate ECU4/10. No close relationship was observed between mAb binding and variations in antigenically critical residues for other isolates. Some unique amino acid changes located in/near antigenic sites were identified on virus capsid structural proteins by sequence analysis. On the contrary, the mAb reactivity with those isolates did not show any significant reduction. The isolate ETH 39/2009 demonstrated high antigenic variation by sequence analysis, but only one mAb binding reduction was detected. One possible explanation may be that although the amino acid mutation occurred in/near antigenic sites, residues are not located in the central part of epitopes. Thus the binding affinity is not affected significantly. Experiments of antibody-antigen binding revealed that the amino acids in the central part of the epitope–antibody site make the majority of the contribution to the antibody–antigen interaction . A second possible explanation may be that despite the amino acid changing, the protein still folded into the similar 3D capsid structure. Thus the antibody binding was not reduced.