Natural T-cell responses generally lack the potency to eradicate cancer. Enhanced affinity T-cell receptors (TCRs) provide an ideal approach to target cancer cells, with emerging clinical data showing significant promise. Nevertheless, the risk of off target reactivity remains a key concern, as exemplified in a recent clinical report describing fatal cardiac toxicity, following administration of MAGE-A3 specific TCR-engineered T-cells, mediated through cross-reactivity with an unrelated epitope from the Titin protein presented on cardiac tissue. Here, we investigated the structural mechanism enabling TCR cross-recognition of MAGE-A3 and Titin and applied the resulting data to rationally design mutants with improved antigen discrimination, providing a proof-of-concept strategy for altering the fine specificity of a TCR towards an intended target antigen. This study represents the first example of direct molecular mimicry leading to clinically relevant fatal toxicity, mediated by a modified enhanced affinity TCR designed for cancer immunotherapy. Furthermore, these data demonstrate that self-antigens that are expressed at high levels on healthy tissue should be treated with extreme caution when designing immuno-therapeutics.