Sabrina Kraus

Human cytomegalovirus (HCMV) reactivation poses a substantial risk to patients receiving tranplants. Effective risk stratification and vaccine development is hampered by a lack of HCMV-derived immunogenic peptides in patients with common HLA-A∗03:01 and HLA-B∗15:01 haplotypes. This study aimed to discover novel HCMV immunogenic peptides for these haplotypes by combining ribosome sequencing (Ribo-seq) and mass spectrometry with state-of-the-art computational tools, Peptide-PRISM and Probabilistic Inference of Codon Activities by an EM Algorithm. Furthermore, using machine learning, an algorithm was developed to predict immunogenicity based on translational activity, binding affinity, and peptide localization within small open reading frames to identify the most promising peptides for in vitro validation. Immunogenicity of these peptides was subsequently tested by analyzing peptide-specific T-cell responses of HCMV-seropositive and -seronegative healthy donors as well as patients with transplants. This resulted in the direct identification of 3 canonical and 1 cryptic HLA-A∗03–restricted immunogenic peptides as well as 5 canonical and 1 cryptic HLA-B∗15–restricted immunogenic peptide, with a specific interferon gamma–positive (IFN-γ+)/CD8+ T-cell response of ≥0.02%. High T-cell responses were detected against 2 HLA-A∗03–restricted and 3 HLA-B∗15–restricted canonical peptides with frequencies of up to 8.77% IFN-γ+/CD8+ T cells in patients after allogeneic stem cell transplantation. Therefore, our comprehensive strategy establishes a framework for efficient identification of novel immunogenic peptides from both existing and novel Ribo-seq data sets.