Epstein-Barr virus infection is strongly associated with a number of malignancies. The EBV latent membrane protein 2A has been implicated as one of the most attractive candidates for immunotherapy of related malignancies. In previous studies, the T cell epitopes of LMP2A have been identified systematically. However, the epitope-based vaccine generally meets inefficient immunogenicity when used in vivo directly, which could be overcome by combination with appropriate adjuvants. Heat shock protein is a natural chaperon, which is able to activate the classical major histocompatibility complex class I antigen-processing pathway (cross-presentation). In this study, a minigene encoding LMP2A(356-364) (FLYALALLL) was genetically fused to the carboxy-terminal of mycobacterial heat shock protein 70. The epitope fusion protein was expressed and purified, and the cross-presentation of LMP2A(356-364) by monocyte-derived dendritic cells pulsed with the epitope fusion protein was evaluated. Results showed that the epitope fusion protein-pulsed mDCs were much more efficient than the single peptide-pulsed mDCs on CTL activation. Immunization of HLA-A2.1 transgenic mice with MtHsp70-LMP2A(356-364) generated peptide specific CTL more effectively than a single peptide plus incomplete Freund's adjuvant (IFA). Growth of LMP2A expressing B16 melanoma tumor cells was suppressed in the vaccinated groups. Our results suggested that MtHsp70-LMP2A(356-364) fusion protein was more effective than the CD8(+) T cell epitope alone on anti-tumor immunity. As a result, the MtHsp70-LMP2A(356-364) fusion protein is considered to be a promising candidate vaccine for EBV related malignancies.