BACKGROUND: A three-antigen DNA-prime/chimpanzee adenovirus 63 (ChAd63) boost vaccine containing pre-erythrocytic Plasmodium falciparum (Pf) circumsporozoite protein (CSP), Pf apical membrane antigen-1 (AMA1) and malaria multiple epitopes (ME) fused to Pf thrombospondin-related adhesion protein (ME-TRAP) elicited higher vaccine efficacy (VE) in an open label, randomized Phase 1 trial against controlled human malaria infection (CHMI) than the two-antigen vaccine DNA/Human Adenovirus 5 (HuAd5) containing CSP and AMA1. The objective of this follow-up study was to determine whether responses to CSP, AMA1 or TRAP MHC Class I-restricted epitopes were associated with VE. METHODOLOGY: Protected (n = 6) and non-protected participants (n = 26) were screened in FluoroSpot interferon gamma (IFN-γ) and Granzyme B (GzB) assays using antigen-specific 15mer peptide subpools spanning CSP (n = 9 subpools), AMA1 (n = 12 subpools), and TRAP (n = 11 subpools). Individual antigen-specific 15mers in the subpools with strong responses were then deconvoluted, evaluated for activities, and MHC Class I-restricted epitopes within the active 15mers were predicted using NetMHCpan algorithms. The predicted epitopes were synthesized and evaluated in the FluoroSpot IFN-γ and GzB assays. RESULTS: Protected and some non-protected participants had similar responses to individual antigen-specific peptide subpools, which did not distinguish only protected participants. However, deconvoluted antigen-specific positive subpools with high magnitudes of responses revealed individual 15mer peptides containing specific and/or predicted MHC Class I (HLA) epitopes. Responses to epitopes were either IFN-γ-only, IFN-γ and GzB, or GzB-only. Due to limitation of cells, most of the analysis concentrated on the identification of protection associated AMA1 epitopes, since most of the predominant pool specific responses were generated against AMA1 15mer subpools. Furthermore, we previously identified protection associated HLA class I-restricted epitopes in a previous gene-based vaccine trial. Seven predicted minimal epitopes in AMA1 were synthesized and upon testing, five recalled responses from protected participants confirming their possible contribution and association with protection, and two recalled responses from non-protected participants. Two protection-associated epitopes were promiscuous and may have also contributed to protection by recognition of different HLA alleles. In addition, strongly positive antigen-specific 15mers identified within active antigen-specific subpools contained 39 predicted but not tested epitopes were identified in CSP, AMA1 and TRAP. Finally, some non-protected individuals recognized HLA-matched protection-associated minimal epitopes and we discuss possible reasons. Other factors such as HLA allele fine specificity or interaction between other HLA alleles in same individual may also influence protective efficacy. CONCLUSIONS: This integrated approach using immunoassays and bioinformatics identified and confirmed AMA1-MHC Class I-restricted epitopes and a list of predicted additional epitopes which could be evaluated in future studies to assess possible association with protection against CHMI in the Phase 1 trial participants. The results suggest that identification of protection-associated epitopes within malaria antigens is feasible and can help design potent next generation multi-antigen, multi-epitope malaria vaccines for a genetically diverse population and to develop robust assays to measure protective cellular immunity against pre-erythrocytic stages of malaria. This approach can be used to develop vaccines for other novel emerging infectious disease pathogens.