The Use of Algorithms with a Novel Score Standardization Method for the Selection of Candidate Immunogenic HLA-Specific Peptides Led to the Identification of New Immunogenic HLA-A*0201 Cytomegalovirus Epitopes.

The identification of HLA class I restricted epitopes within immunogenic proteins that are capable of stimulating memory T lymphocytes is an important step for both clinical vaccine and adoptive immunotherapy. Despite the strong immunogenicity demonstrated by specific HLA restricted peptides, the peptide’s affinity to a specific HLA antigen may vary among individuals. Therefore, it is important to identify more than one immune determinant for each HLA antigen of interest. Two computer algorithms, one by Parker and another by Rammensee, are widely used in immunogenetic studies in order to predict which peptides bind best to specific HLA molecules. The use of these algorithms represents the first step in screening proteins. Parker’s algorithm ranks potential peptide according to the predictive half-time disassociation of peptide/MHC complexes. Rammensee’s algorithm ranks the peptides based on the presence of primary and secondary HLA-binding anchor residues. Starting with the sequence of Cytomegalovirus (CMV) phosphorylated matrix protein 65 (pp65; SwissProt ID: P06725), we derived from each algorithm a list of 250 nonamer (9-mer) peptides predicted to bind to the HLA-A*0201 allele. Standardizing the two groups of scores, corresponding to the endpoints of the stated interval, by the following formula: [(peptide score - mean of scores)/ standard deviation of scores], we obtained pure values that are independent of the unit of measurement, therefore homogeneous and thus comparable. This allowed us to link each peptide to a pair of values, one for each algorithm. The values derived from standardized Parker scores (X axis) were plotted against the values derived from standardized Rammensee scores (Y axis). We represented final results by a graph. All peptides having positive values with both algorithms and included in the first quadrant of the dot-scatter graph (double positive quadrant (+/+)) were likely to be highly immunogenic epitopes. Those peptides were included in a list of potential peptides to screen. In the presence of homoschedasticity, we selected only peptides showing mean values greater than 1 (among those having both the mean value greater than 0 and the standard deviation value greater than 1). Nine out of the 250 peptides were thus chosen within the CMV pp65 and their immunogenicity analyzed by ex vivo stimulation of PBMC and measured at the transcriptional level using the production of IFN-γ mRNA. Four out of nine peptides stimulated PBMCs from all HLA-A*0201 restricted CMV-seropositive subjects tested (p values: peptide #5, 0.011032; peptide #7, 0.015225; peptide #8, 0.025221; and peptide #9, 0.007302). The intracellular IFN-γ protein production by ex vivo peptide stimulation ranged from 0.14 to 0.62% compared to 0.03% for non stimulated cells. Moreover, each of the four selected peptide, either alone or in pool, was able to reactivate in vitro a strong immune T cell memory response, as measured by cytokine protein release and target cell lysis. The identification of multiple HLA class I restricted epitopes will allow the simultaneous administration of peptides with the ability to produce CMV-specific CTLs in patients bearing the same HLA types. This represents an extraordinary advantage for CMV adoptive immune therapy or vaccination.