Wild-type or immunoevasive antigens can drive weak CD8+T-cell responses against both dominant and subdominant epitopes during gene-based vaccination. For many antigens, fusion to ubiquitin (Ub) to target them to the proteasome circumvents this problem. Although this procedure works in most cases, for one subset of antigens, Ub fusion does not improve immune responses. To determine why these failures occur, we have evaluated in detail the 'rules' for proteasome targeting that have been applied in mammalian vaccine studies, but that were actually defined in yeast systems. To do this, we fused a series of engineered Ub genes to green fluorescent protein (GFP) and tested their ability to target GFP to the proteasome for enhanced antigen processing and CD8+ T-cell responses. Here we demonstrate that Ub fusion mediates enhanced CD8+ responses by proteasome targeting rather than by enhancing protein translation. We also show that several of the yeast-defined Ub constructs failed to target the proteasome in mammalian cells and likewise failed to enhance transgene-specific CD8+ T-cell responses in mice. In contrast, when mammalian-optimized constructs were applied to target the influenza virus nucleoprotein, CD8+ responses were enhanced against its refractory subdominant epitope in mice. This work demonstrates that Ub fusion has efficacy to enhance CD8+ responses, especially against subdominant antigen epitopes, provided constructs are optimized for mammalian use.