Does Similarity Between Self and Non-Self Influence Immunogenicity?

What is the impact of CHO (Chinese Hamster Ovary) Host Cell Proteins  in Therapeutic Protein Formulations?
New information emerging from a collaboration between Chris Bailey-Kellog, Andres Gutierrez, Leonard Moise, Frances Terry, Bill Martin and Anne De Groot (Dartmouth College, University of Rhode Island, and EpiVax, Inc.)
We thought that the publication of the CHO genome provided a unique opportunity to examine the importance of self/non-self differences in similar proteins. The reason that this is important is that we observed that immune response to FVIII and to thyroid hormone receptor in hemophilia and Grave’s disease mouse models targets the epitopes that are different between human and mouse. While not surprising, this does suggest that these uniquely different epitopes may be the ones that drive immune responses to CHO. Thus we analyzed proteins from three CHO datasets, described here (and in great detail in a paper to be published soon) so as to explore their T cell epitope abundance. These datasets consisted of expressed, putatively secreted, and experimentally determined CHO contaminating proteins.

Epitope Network comparing CHO and Self (Human) proteins.

After arriving at a final subset of constitutively expressed CHO proteins (1757 in total), we used our epitope prediction tool, EpiMatrix, to identify CHO proteins that had the potential for immunogenicity. We also compared the epitopes, as illustrated in the diagram, to common human “self” proteins. These preliminary results will be published as part of the proceedings of ICIW.

In our preliminary analysis, 26% of the 1757 proteins analyzed were above our threshold for likely immunogens, containing an abundance of predicted epitopes. We then compared these potential immunogens for conservation with the human genome. Some example results are illustrated the figure above; what is remarkable the number of epitopes that are not cross reactive with the human genome (represented by grey boxes with no attachments to similar epitopes (triangles) within the human genome).

The epitope network illustrated here should be reassuring except where the connections between CHO and non-CHO are absent; those epitopes appear as clusters that are present in one protein, but are not connected to any other protein. The immunogenicity of the ‘dissimilar’ epitope may also be influenced by whether the self-epitopes it is cross-reactive with are very commonly expressed. Thus we are now in the process of weighting the epitopes by frequency that they occur in the human genome, and by prevalence of the human host protein.

In summary, we are a long ways down the road in the important process of uncovering the potential for HCPs to contribute to unwanted immunogenicity. We’re sharing our thoughts in the interest of getting your feedback. Is this of interest? Should we post the results online? Let us know.