Gaiha, Gaurav DGayton, Alton Charles2026-06-0920262026-06-052026Gayton, Alton Charles. 2026. Rational Design of Therapeutic T cell-Inducing Vaccines for Human Papillomavirus. Doctoral Dissertation, Harvard University Graduate School of Arts and Sciences.32701666https://dash.harvard.edu/handle/1/42740351Current prophylactic human papillomavirus (HPV) vaccines are highly effective at preventing viral acquisition and cervical cancer by inducing neutralizing antibodies. However, they provide no therapeutic benefit for individuals already infected with HPV. Consequently, women with HPV-positive precancerous lesions often rely on surgical removal of cervical tissue, an intervention that is inconsistently available in low- and middle-income countries (LMICs). Existing therapeutic vaccine strategies have primarily aimed to induce CD8+ T cells and target the oncogenic HPV proteins E6 and E7. However, these proteins are most prominently expressed during later stages of HPV-driven malignancy, which limits the ability of such vaccines to clear early-stage infections. In contrast, several HPV early proteins, including E1 and E2, are continuously expressed during early infection and are substantially larger than E6 and E7, potentially providing a broader set of immunogenic targets for CD8+ T cells. We therefore hypothesized that therapeutic vaccines targeting multiple HPV proteins expressed across different stages of infection could induce a more comprehensive antiviral immune response. To test this concept, we developed rationally designed therapeutic T cell vaccines encoding epitopes from multiple HPV16 and HPV18 early proteins, including E1, E2, E5, E6, and E7. To identify optimal targets, we applied MUNIS, a deep learning-based epitope-prediction model, to define immunogenic regions across the HPV16 and HPV18 proteins. Predicted epitopes restricted by HLA class I alleles common in sub-Saharan Africa and those covering approximately 99% of global HLA diversity were prioritized. Non-immunogenic regions were excluded to reduce cassette size and improve stability. The resulting sequences were assembled into HPV16-only, HPV18-only, and bivalent HPV16/18 constructs and formulated as mRNA-LNP and circRNA-LNP vaccines. Immunogenicity was evaluated in C57BL/6 mice as well as human HLA-A02 and HLA-B07 knock-in mouse models. Across all models, vaccination elicited broad and robust CD8+ T-cell responses, particularly against E1. In addition, vaccine-induced CD8+ T cells recognized several previously unreported MUNIS-predicted epitopes restricted by HLA-A02 and HLA-B07, highlighting the potential of artificial intelligence to expand epitope discovery even for well-characterized HLA alleles. Therapeutic efficacy was evaluated using established and newly developed tumor models designed to simulate distinct stages of HPV-associated disease, including (1) clearance of cervical cancer (E6/E7-expressing tumors), (2) prevention of HPV-positive tumor formation (tumors expressing E1, E2, E5, E6, and E7), and (3) clearance of early cervical intraepithelial neoplasia (tumors expressing E1, E2, and E5). In all three models, vaccination produced robust CD8+ T cell-mediated antitumor activity. Together, these findings demonstrate that AI-guided design of multi-antigen therapeutic vaccines can generate broad CD8+ T cell immunity against HPV proteins expressed across multiple stages of infection and malignant transformation. This strategy may provide a promising therapeutic approach for clearing HPV infection and preventing cervical cancer, particularly in high-incidence LMIC settings.application/pdfenCancerCervicalHPVTherapeuticsVaccineVirologyImmunologyThesis or Dissertation2026-06-090000-0002-2411-0621