MAGE A3, which belongs to the family of cancer-testis antigens, is an attractive target for adoptive therapy given its reactivation in various tumors and limited expression in normal tissues. We developed an affinity-enhanced T cell receptor (TCR) directed to a human leukocyte antigen (HLA)–A*01–restricted MAGE A3 antigen (EVDPIGHLY) for use in adoptive therapy. Extensive preclinical investigations revealed no off-target antigen recognition concerns; nonetheless, administration to patients of T cells expressing the affinity-enhanced MAGE A3 TCR resulted in a serious adverse event (SAE) and fatal toxicity against cardiac tissue. We present a description of the preclinical in vitro functional analysis of the MAGE A3 TCR, which failed to reveal any evidence of off-target activity, and a full analysis of the post-SAE in vitro investigations, which reveal cross-recognition of an off-target peptide. Using an amino acid scanning approach, a peptide from the muscle protein Titin (ESDPIVAQY) was identified as an alternative target for the MAGE A3 TCR and the most likely cause of in vivo toxicity. These results demonstrate that affinity-enhanced TCRs have considerable effector functions in vivo and highlight the potential safety concerns for TCR-engineered T cells. Strategies such as peptide scanning and the use of more complex cell cultures are recommended in preclinical studies to mitigate the risk of off-target toxicity in future clinical investigations.