The CTLA4-Ig therapeutics abatacept and belatacept inhibit CD28-mediated T cell activation by binding CD80 (B7-1) and CD86 (B7-2) co-stimulatory ligands. Both compounds preferentially bind CD80, yet CD86 has been implicated as the dominant co-stimulatory ligand. Using directed evolution methods, novel CTLA4-Ig variants were created with selective CD86 binding affinity, a property that confers increased immunosuppressive potency and potentially improved efficacy and safety profiles. Relative to abatacept (wild-type CTLA4-Ig), ASP2408 and ASP2409 have 83-fold and 220-fold enhanced binding affinity to CD86 while retaining 1.5-fold and 5.6-fold enhanced binding affinity to CD80, respectively. Improvements in CD86 binding affinity correlates with increased immunosuppressive potency in vitro and in vivo. Our results highlight the power of directed evolution methods to obtain non-intuitive protein engineering solutions and represent the first examples of CD86-selective CTLA4-Ig compounds that have entered clinical trials.