Little is known about the repertoire of MAGE-A3 CD4 + T-cell epitopes recognized in vivo by neoplastic patients and how antigen processing influences epitope formation. Here, we first show that MAGE-A3-specific CD4 + T cells are present in the blood of advanced melanoma patients. MAGE-A3 111-125 , , and MAGE-A3 281-300 were recognized by 7, 6, and 5 of the 11 patients tested, respectively. MAGE-A3 146-160 and MAGE-A3 171-185 were also recognized in two and one cases, whereas no recognition of MAGE-A3 161-175 and MAGE-A3 243-258 was observed. Cytokines produced were mainly interleukin 5 and/or granulocyte macrophage colony-stimulating factor, suggesting impairment of productive polarized Th1 responses. Secondly, proteases inhibitors were used to modulate in vitro the recognition by CD4 + T-cells clones of dendritic cells loaded with MAGE-A3-expressing cell lysates. We found that formation of MAGE-A3 111-125 depended on both leupeptin-sensitive and pepstatin-sensitive proteases. In contrast, we found that MAGE-A3 161-175 , which was never recognized ex vivo, was formed by leupeptin but destroyed by pepstatin-sensitive proteases. Collectively, our results show that (a) anti-MAGE-A3 CD4 + T-cell immunity develops in vivo in neoplastic patients and is focused toward immunodominant epitopes, (b) the response in advanced disease is skewed toward a Th2 type, and (c) endosomal/lysosomal proteases in dendritic cells influence the repertoire of the epitopes recognized. [Cancer Res 2008; 68(5):1555-62]