PREDICTING AND REDUCING IMMUNOGENICITY
As discussed the mechanisms leading to antibody induction by therapeutic proteins are still not completely understood. As a consequence it is impossible based on our current knowledge to fully predict the immunogenicity of a new product in patients. For nonhuman proteins which induce the classical immune response, the level of nonself is a relative predictor of an immune response. However, it is not an absolute predictor. Sometimes a single amino acid change is sufficient to make a self protein highly immunogenic. With other proteins substantial diver-gence from the natural sequence has no effect. For foreign proteins a number of in vitro stimulation and binding tests and computational models are adver-tised as predictors of immunogenicity. However, all these tests have their limitations. T-cell proliferation assays, for example, have the drawback that many antibodies are capable of inducing some level of T-cell activation or inhibit cell proliferation. The computa-tional algorithms which predict binding of antigens toHLA class II only give limited information on the interaction of the proteins with the immune system and also under-detect epitopes (Stevanovic, 2005). These limitations are also evident when these assays or algorithms are used to reduce immunogenicity: there is hardly any convincing evidence of a clinically relevant reduction of antibody induction.
For human homologues the best predictor of immunogenicity is the presence of aggregates and to a minor degree the presence of impurities. Thus, the quality of the therapeutic protein and its formulation are important factors. There is also evidence about immunogenicity introduced by a change in formula-tion and a reduction in immunogenicity by avoiding aggregation and improving purification and formula-tion (see above).
Although animal studies are helpful in obtaining control sera and may provide insight in the possible clinical effects of immunogenicity, they are not very good predictors of immunogenicity in patients. All proteins, including the human homologues, will be in principle immunogenic in animals. Sometimes animal studies may help to study the relative immunogeni-city of different products or formulations, although their predictive value for the clinic is questionable. Even monkeys studies do not completely predict immunogenicity in patients. Some products which are immunogenic in nonhuman primates do not induce antibodies in humans and vice versa.
The animal model to study the factors important for the breaking of tolerance are transgenic mice carrying the gene for the human protein (Hermeling et al., 2005). These animals have an immune tolerance comparable to the immune tolerance in patients. These animals have also been successfully used to identify new epitopes in modified products. Although transgenic animals have been proven to be important scientific tools, the experience is still too limited to claim they can serve to completely predict the human response.
Several strategies are being applied to reduce im-munogenicity besides changing the amino acid sequence of the product. Linking proteins to polymers such as polyethylene glycol and low-molecular-weight dextran reduces immunogenicity. However, these modifications also make the mole-cules less active, necessitating higher doses. This and their increased half-life extend their exposure to the immune system, which can increase the immuno-genic potential. Another approach is to reduce the immunogenic response by immunosuppressive treat-ments. In addition, tolerance induction is being applied, e.g., in hemophilia patients with antibodies to factor VIII.