Autoreactive cytotoxic CD8 T-cells (CTLs) play a key pathogenic role in the destruction of insulin-producing beta-cells resulting in type 1 diabetes. However, knowledge regarding their targets is limited, restricting the ability to monitor the course of the disease and immune interventions. In a multi-step discovery process to identify novel CTL epitopes in human preproinsulin (PPI), PPI was digested with purified human proteasomes, and resulting COOH-fragments aligned with algorithm-predicted HLA-binding peptides to yield nine potential HLA-A1, -A2, -A3 or -B7-restricted candidates. An UV-exchange method allowed the generation of a repertoire of multimers including low-affinity HLA-binding peptides. These were labeled with quantum dot-fluorochromes and encoded in a combinatorial fashion, allowing parallel and sensitive detection of specific, low-avidity T-cells. Significantly increased frequencies of T-cells against four novel PPI epitopes (PPI_4–13/B7, PPI_29–38/A2, PPI_76–84/A3 and PPI_79–88/A3) were detected in stored blood of patients with recent onset diabetes but not in controls. Changes in frequencies of circulating CD8 T-cells against these novel epitopes were detected in blood of islet graft recipients at different time points after transplantation, which correlated with clinical outcome. In conclusion, our novel strategy involving a sensitive multiplex detection technology and requiring minimal volumes of stored blood represents a major improvement in the direct ex-vivo characterization and enumeration of immune cells in the pathogenesis of type 1 diabetes.