Non-conserved framework surface residues, which were not deemed too close to CDRs were replaced, but experimental analyses revealed that of these, 6 positions in Vl and 10 positions in Vh substantially affected antibody interaction with the antigen. For that reason a compromise between potential immunogenicity and retained binding specificity had to be made. Finally, 13 murine residues were replaced in the moScFv V5B2 to prepare a huScFv V5B2, which retained the significant ability to discriminate between CJD-affected and normal brain tissue. Nevertheless, the amino acid sequence of huScFv V5B2 shows high similarity with the human heavy chain subgroup III and light chain k subgroup I consensus sequences. In the process of humanization a computer model of antibody variable domains is often built to help design the humanized form. It is used for prediction of possible influence of mutations on CDR conformations, which usually results in loss of antibody binding BAY 73-4506 affinity or even specificity. Several reports showed that analysis of a computer model actually helped to avoid problems with the affinity reduction, which is particularly typical for CD-grafting. Moreover, it was demonstrated that during humanization antibody affinity could even be improved when the humanized form is carefully designed on the basis of a precise analysis of structural models. In our case, the structural model of the Fv V5B2 helped to determine framework surface residues of the V5B2, but it did not predict negative impact of several replacements we have introduced. It was however reasonable to expect that not all planned mutations could be introduced into variable domains without disturbing the structure of the antigen-binding site and influencing the binding. For that reason, several intermediate variants were prepared and tested for antigen-binding activity. A few resurfaced scFvs have been reported in the literature, generally containing from six to ten replacements. Any additional mutation usually resulted in reduced binding activity. It was also shown that even a single mutation in the antibody framework can improve or reduce the expression yield or binding affinity of a scFv tremendously. In our experiments, Western blot analyses indicated that humanized V5B2 scFv recognized the same epitope on PrP as the parent V5B2 mAb. When huScFv was assayed by IHC, it labelled less kuru plaque-like PrPSc deposits than V5B2 and failed to label the synaptic pattern of PrPSc deposition, which was clearly visualized by whole V5B2 mAb. This observation was attributed to expected reduced affinity of scFvs that hindered the detection of fine synaptic pattern and small plaques. Our IHC experiment clearly demonstrated that both murine and humanized form of scFv retained the ability to label PrPSc deposits specifically, although less potently, which is in agreement with results obtained by ELISA and immunoblotting. Even though several antibodies have been resurfaced in the last decade, their immunogenicity remains undetermined.